Home

VisuaLyzer User Manual

1. e To import a GraphML file select Import from the File menu This will display a standard Windows file dialog like the one in Figure 14 page 22 Select GraphML files in the Files of type box to have these files displayed When you have found the file you want to import either double click or single click on it for it to be displayed in the File name box and then click the Open button Your data will be displayed as a network graph 25 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved DyNetML DyNetML is another XML derived language that provides means to express rich social network data As data interchange format it improves compatibility of analysis and visualization tools DyNetML provides an extensible facility for linking anthropological process description and other data with social networks DyNetML has been implemented by the CASOS group at Carnegie Mellon University Further information about DyNetML its history background and existing parsing and conversion software can be found at http reports archive adm cs cmu edu anon isri2004 abstracts 04 105 html An example of simple 3 node graph above is produced below lt xml version 1 0 encoding UTF 8 gt lt DynamicNetwork gt lt MetaMatrix timePeriod 1 30 2007 11 50 52 AM gt lt nodes gt lt nodeset id VisuaLyzer type agent gt lt node id Manuan gt lt node gt lt node id Allen gt lt
2. For directed graphs we also compute node in degree the number of edges received by the node and the out degree the number of edges initiated by the node Loops if any are counted twice The out degree indicates the role of an actor node as a source of ties in a directed graph This is expressed as the sum of the connections from the actor to others e g actor 1 sends information to four others Out degree is usually a measure of how influential the actor may be For example the first set of actors 18 15 21 and 10 have high out degrees meaning they give advice to a host of others By contrast in degree indicates the role of an actor node as sinks or receivers of information This is expressed as the sum of the connections to each node that is how many other actors send information or have ties to a specific actor of interest The in degree of an actor may be an index of prestige because other actors want to be known by the actor hence they send information ties But these high in degree actors may also be burdened by information overload or noise and interference due to contradictory messages from different sources For example actors 2 18 and 21 have the most in degrees indicating that they are the ones consulted the most for advice Standardized normalized indices allow comparison of degree centrality scores across other networks of different sizes Normalized degree scores are a ratio of the degree out or
3. There are two quick ways to disable and later re enable nodes e Select one or more nodes and press the Delete key e Hold down the CTRL key and click on a node You can also select Enable nodes from the Edit menu to display the Enable Nodes screen Figure 45 Enable Nodes gums C Alexis Alyssa Christopher Emily Group_001 Hannah Jacob Joseph Lauren Matthew C Nicholas Figure 45 Enable nodes screen By default nodes that are checked are enabled To disable the nodes uncheck the box next to that node s name Click the Enable Selected button to enable all selected nodes and disable all other nodes Click the Disable Selected button to disable all selected nodes and enable all other nodes Clicking the Enable All or Disable All buttons will enable or disable all nodes in the graph respectively Disabled nodes can be moved manually but are ignored by the automatic layout routines including AutoOrganize You can also edit disabled nodes and links 57 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Grouping Nodes VisuaLyzer allows you to group two or more nodes together into a single node To do this select the nodes you want to group together then right click on the graph and select Group nodes from the pop up menu or press the F8 key To ungroup the nodes in a group select the group and then select Ungroup from the right click pop up menu If y
4. efe Visual yzer 2 0 Default Style Default 3D_Cube txt DER File Edit Layout View Query Analysis Reasoning Help SAB BAS A OR BOOM E p gai peee iode_o6 awa l n nn Pa h Nece_o3 yn e h ae Node_01 niode_07 i Nodes 8 Links 12 Use Dbl click to add or edit a node link right button add linkfaccess menu left button drag select disable node s Figure 37 Best fit for cube graph Node Position You can also position or align selected nodes manually by right clicking on the graph screen and selecting Node s position gt Align This will give you a set of options to align the nodes horizontally or vertically as shown in the figure below 47 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Daniel Qa Y f _A Sarah Abigail David Alignment amantha Horizontal Vertical e O no change O no change Elizabeth left most O top most O center O center O right most O bottom most O space equally 5 space equally center in screen O center in screen John Jacob tauren f o Alexis ds Matthew Figure 38 Selected nodes position aligned General Options Visual yzer has other graph wide display options as well These options can be accessed and set by e Pressing the Display Properties button e Pressing the F4 key e Selecting Edit gt Display Properties gt View Edit from the main m
5. foe hode p NNode_013 Nodes 15 Links 16 Use Dbl click to add or edit a nodejlink right button add linkfaccess menu left button drag select disable node s Figure 6 Adding a node The new node Node_015 is created and also selected A node is selected if it is framed as is node 015 Adding Links To link two nodes together right click and hold on the first node then drag the line and cursor to another node and release the right click button Fig 6 To link an existing node and create a new one at the same time release the right button at the desired position of the new node Links can also be created by selecting two nodes and pressing the F7 key Figure 7 shows the creation of a new link between the new Node_015 and Node_001 13 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved de Visual yzer 2 0 Default Style Default Untitled File Edit Layout View Query Analysis Reasoning Help 1209 BMS AACR BOOMS LPS REMY HD Node_006 ARE Click the display ree A a properties button A Fi Foa A ae sa N AD Node _003 j F mai p A pS lode_001 Qhode_008 j J j Y S Node_007 Node_o09 r A Y itede_o12 4 Node_010 hode p fiode 013 Nodes 15 Links 17 2 nodes selected Figure 7 Adding a new link The new link is created and selected and the two nodes it connects are also se
6. Using the provided example file CakeHouse_2 mode eng this will display a 2 mode network as it is shown in Figure 35 de Visual yzer 2 0 Default Style Internal CakeHouse_2 mode eng DER AE AD NORKA Tha Shining The XS lack Nicholson g Batman E QS Stanley Kubrick A Clockwork Orange s movie actor Nodes 5 Links 4 Use Dbl click to add or edit a node link right button add linkfaccess menu left button drag select disable node s Figure 35 Simple bipartite graph 45 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Best Fit As you move nodes and links around you can automatically center them and spread them out for easier readability by using the Best Fit feature To use this feature either select Best Fit from the Edit menu or click the Best Fit button For example before using Best Fit a graph might look like Figure 36 f Visual yzer 2 0 Default Style Default 3D_Cube txt File Edit Layout View Query Analysis Reasoning Help iets BAS os Nodes 8 Links 12 Use Dbl click to add or edit a nodejlink right button add link access menu left button drag select disable node s Figure 36 Cube graph 46 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved After clicking the Best Fit button the graph would look like Figure 37
7. 4 James 4 Sarah gt a Abigail Qo j P a A Alyssa A ce fit Sh Christopher Joseph f 4 Jacob Ja P F X x pl Alexis Matthew Nodes 18 Links 17 Use Dbl click to add or edit a node link right button add linkfaccess menu left button drag select disable node s oa N Figure 51 Adding meaning to a basic graph The process for changing the first graph to look like the second is described below Node Label To base node labels upon the value of one of the nodes attributes e select Query gt Attribute Based gt Label e orclick the Attribute Based Query button and select the Label tab This will display the Label tab of the Attribute Based screen Fig 52 64 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Attribute based ee Label Color Shape Size Link Color Label nodes with the value of Figure 52 Attribute base selection screen The default is to label nodes with the value of their Label property but you can have the nodes labeled with the value of any of their attributes Select the attribute whose value you want used as a label in the drop down list box at the top of this screen and then click the Apply button Node Color To have the node color based upon the value of one of the nodes attributes e select Query gt Attribute Based gt Color e orclick the Attrib
8. 5 the converse of the relation R is converse R x yl y R x 6 the identity relation relative unit which is the set of pairs x ylx y is id 7 a selector p n j is the set x y for every x1 x2 xn x p x1 x2 x 1 y x 1 xnj These operations are available within the Reasoning gt Arbitrary Relations gt Relation Combinators menu Note the hints at the bottom of the display An individual selector and any expressions involving inequalities are not depicted in VisuaLyzer as a network but as a set expression Generally every set of directed edges ordered pairs is a relation but a relation is not necessarily a set of directed edges The identity relation could be envisioned as a set of self loops on an infinite set of nodes and depicted as a single self loop on a node with a name as variable or not available NA Even if R is a finite set of edges the constructive complement R is an infinite set complement R can always be finitely described as a collection of perhaps existentially quantified equations and universally quantified inequations in disjunctive normal form Thus complement R differs from the graph complement The relation complement represents the negation of all the knowledge in the relation The graph complement is the set of edges that are not in the original network For a network g the graph complement is complete g complement g if complete g is the complete graph containing g These primitives can
9. Copyright 2005 2007 Medical Decision Logic Inc All rights reserved After selecting New Random from the File menu you will see the dialog box below Random Layout Enter a number of links or a percent of total possible links here Select the number of nodes for your network here Number of nodes Number of links Click the OK button Figure 2 Creating a random graph Enter the number of nodes and either the number or percentage of links click the OK button to see your new graph The graph shown below was created with 24 nodes and 138 links Visual yzer 2 0 Default Style Default Untitled File Edit Layout View Query Analysis Reasoning Help 12903 AHS AOR AG lode_005 A Node _001 Node _004 Click spring embedding layout button SS eh N N77 a 20 Node_D12 SN gt ina SC Node_008 Nodes 24 Links 138 ERASE STE EEES EN Figure 3 A random graph 10 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Using Graph Layouts You have just created your first graph in VisuaLyzer However the randomness of the arrangement of the nodes makes it difficult to get a good overall view of the network structure so next we will use automated layout routines to arrange the graph Spring embedding layout To organize your graph using a spring embedding layout select Spring Em
10. VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Adjacene y MIA i saat vat E eli A a i Oh EE Satu le ad 37 Displav OTI i EEE E E A EE A 38 CE CUM is a di E a e 38 Graph Layo t en easa TE aE A ERS disco 39 A 39 Movme O e e e e tbn aR 39 Spring Embedding Layout tii dire 40 Coca E e elo 40 Radial LAU A A rate Dias ceed od late A sete geld 40 e D 6 i Rene pee Be oP ne O OO AE or EOE REE 40 N mode AV OUE errn dis 44 Best Piti stacsa seas anes suntan vs ds 46 NOA SHENG ges gaasttal te o dd o lo 47 Genera OOS SI as dios 48 Background Color vivi Ai 49 Background Marui A dt 50 SUT GOL aa vd hase is caw asa eh ecg a alc ga esr eal alah AD SN ea deh I aoa neal 50 A OR 50 Selection Frame Color ai SEEKER 50 Cluster Frame Color A ai 50 POSH Tre FATT COIR sca oe Saat a E a eh ee 50 SS VAIN Sass aspy els sr es pa a dav sures A en aes 51 SHOW IMa GES iii dais A At asiendo aes 51 a A E seeeeane 51 Show e olla sie e ees 51 A sates sales e cu otc Ses EE E TE E A 51 SHOW LOOP nai is 51 Show Links ici iia a cert des 51 Show LNK RO 51 Show Link Type relation ceire a a a e E a Atak E S 52 Show Link Cabeleira eaea e T idas 52 Show Link Weight hacen i ee a Ai ais 52 AMOS AVE rA icin ae ig alee E i i A Tae OE E E E a 52 AI A A t 52 Sounds Effects a a e e 52 Node options ener eseri naea a Ea e E RS T A E E EAEE A EAA TE asad ee 52 Eo Lo ee ae eer et er e EE A N EART EEE EE 54 SOE UID pcre i
11. You can change the shape used for an attribute value by double clicking on the shape to the left of it and then clicking on a shape from the Select shape screen that is displayed If you choose the icon A from this screen only the node s label will be displayed As in the case of basing color on an attribute the Show Image box on the General tab in the Display Properties selection screen should be un checked Node Size To have node size be based upon the value of one of the nodes attributes e select Query gt Attribute Based gt Size e orclick the Attribute Based Query button and select the Size tab This will display the Size tab of the Attribute Based screen Select the attribute you want used to determine size in the drop down list box at the top of this screen The Legend box will then be filled with a list of the values for that attribute with a number next to each value indicating the size in pixels of the image or shape that will be used for nodes with that value Fig 56 68 VisuaLyzer User Manual v 2 0 Attribute based PS Label Color Shape Size Node attribute AGE Legend Size Value Gba Copyright O 2005 2007 Medical Decision Logic Inc All rights reserved Link Color v Close Figure 56 Attribute selection screen size tab You can change the size used for an attribute value by double clicking on the number to the left of it and then entering a number in the
12. complement R switches true and false while converse R switches object position The Combinators and the Reciprocals are functions from relations into relations The Select Relation box of Figure 60 is exposed with Edit gt Select Relation or Ctrl R The arguments to operations such as Combinators or Reciprocals are selected in the Select Relation box Consider the well known example of David Krackhardt s Hi Tech managers To simply construct the product of the relations reports_to and is_friends_with place check marks next to those two relations in the Select Relation box then perform Reasoning gt Arbitrary Relations gt Relation Combinators gt Absolute Product R S In this instance R represents the relation reports_to and S represents is_friends_with The resulting network those managers that report to a friend appears as a new window Even though these combinators apply in this instance to two relations R and S they can extend to one two or more When they apply to just one relation R the other S defaults to the identity for the selected combinator that is R 0 R 1 R amp di and R id If more than two relations are chosen in the Select Relation box then the combinator is extended in the obvious way The dichotomizer menu choices within Reasoning gt Arbitrary Relations gt Relation Combinators do not all described networks but sometimes produce set expressions that represent specifi
13. menu or click the Circular Layout button Fig 4 This will change our graph to look like Figure 5 ee Visual yzer 2 0 Default Style Default Untitled File Edit Layout Yiew Query Analysis Reasoning Help 1208 ARS QOD HOOMS PPr mA MORA EQ O Node_006 AQ Node_004 N A _LOMNode_005 Double click here ode MS Lh ode _o0z NO XQ Mode_003 a Fi i i ode_001 y Node_008 B i f f Sree OnNode_009 A A TE X dna _ Node_010 O node_011 Qfode_013 Nodes 14 Links 16 Use Dbl click to add or edit a nodeflink right button add linkfaccess menu left button drag select disable node s Figure 5 Circular layout Note that Node_006 is still in the upper left Adding Nodes Next we ll add some additional nodes to our graph To do this simply double click anywhere in the VisuaLyzer window to make a node appear Fig 5 Your screen will then look like Figure 6 with the addition of a new node 15 12 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved de Visual yzer 2 0 Default Style Default Untitled File Edit Layout View Query Analysis Reasoning Help 1298 BMG RA Ow BOOMS 220440 NN BARGE O Node_006 QNode_o14 IN TQ Node_004 ai Y Se W i A Right click here AQ Mode _005 fs A ee sa N 0 Node_003 A K WNode_015 E j Node _007 and drag to here Ove
14. 23 3D Cube graphml If the example file 3D_Cube graphml Fig 23 is merged with the first file all but one of the new nodes Node_06 will match existing nodes based on their names i e both nodes named Node_01 will match and thus only one new node is added However the links of the existing nodes are updated to display the links of the nodes in the new file as well The result looks like the graph shown in Figure 24 File format is not important when adding nodes 33 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved eo OMode_13 ri Bi f Node_19 f T Sa f ia Node_14 O Node_25 is i Node_06 Node_20 f i y t Node_26 Node_15 oz 4 f Node_21 Node_22 j s Ny f Node_28 f Hi Node_29 Figure 24 Addition of existing nodes only adds new links Adding new Node Attributes You can also add new node attributes to an existing graph which is useful if you have re coded data or collected new information on your existing subjects In the previous example using the tree with edge attributes graph Fig 20 page 31 each node had four attributes Age Gender Kids and Marry We can add new node attributes by selecting Add Node Attributes from the File menu then selecting the provided node attribute file Tree with EdgeAttributes AddNodeAttr txt The nodes will now have two additional attributes of Education and YearsOfExperience Note
15. Excel sheets have a header row of variable names which means the real data starts in the second row instead of the first row Finally click OK to start importing data If the data is in the form of an adjacency matrix select the Adjacency Matrix tab Fig 17 Select Excel Sheet Edgelist Adjacency Matrix Workbook C Visualizer Examples Excel Status xls Look adjacency matrix in Sheet3 AdjacencyMatrix Y id labels Look node attributes in EEEN assuming 1 row and 1 column for labels Header row Figure 17 Importing adjacency matrix from Excel 28 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved CSV files comma separated values To import data from a CSV file select Import from the File menu This will display a standard Windows file dialog like the one in Figure 14 Select CSV text in the Files of type box on this screen to have these files displayed When you have found the file you want to import either double click on it or single click on it for it to be displayed in the File name box and then click the Open button The data will be displayed as a series of nodes at the left hand side of the screen Fig 18 The nodes can then be moved and linked to show relationships In Figure 18 Daniel and Alyssa have been moved and linked to show a relationship Attributes can also be imported Daniel was selected and double clicked t
16. Logic Inc All rights reserved Layered Layout Select node attributes to use in Layered Layout Y axis Arrange layers according to value of AGE O according to node degree Sort order Descending Ascending x axis Arrange layer nodes O evenly along X axis O according to value of AGE O according to node degree ased on optimization algoril based timization algorithm min edge crossings Figure 32 Layered Layout setup screen 2 Clicking OK then displays the much more readable graph shown in Figure 33 Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng File Edit Layout View Query Analysis Reasoning Help 1208 BAL AOD BOOMS SP AO WERA BERGE A Abigail a XM y A l a Ss te 44 pi Tie ss A Di Daniel 5 Christopher 5 Joseph Sarah o gt m Mathew A Lauren Nicholas i Elizabeti pava 4 Alexis Jom AR samantha as i AZ AE A Er Nodes 18 Links 17 Use Dbl click to add or edit a nodeflink right button add linkfaccess menu left button drag select disable node s Figure 33 Layered layout graph 3 43 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Nodes can also be displayed along the X axis by node degree number of links or the value of node attributes Displays like this can help you determine if the values
17. The graph used in this example can be found at Examples Binary KrackHighTec eng in the installation folder Select Analysis gt Core and Periphery from the menu This will present a new graph partitioning the network into core and peripheral nodes Fig 80 The user can select the core or peripheral nodes by the clicking on the corresponding Show buttons on the Core Periphery form Once selected the user can apply color shape and other visual properties to distinguish between the two kinds of nodes The associated Core amp Periphery form presents detailed results of the analysis procedure Total and Enabled Nodes This shows that there are 21 nodes in the data and all are enabled available for analysis Core Nodes Number of nodes in the core Periphery Nodes Number of nodes in the periphery Core Periphery Ratio The number of Core Nodes divided by the number of Peripheral Nodes Core Centralization An indicator of fit to the ideal model The higher the coefficient closer to 1 the better the fit is to the ideal model The results from the Core and Periphery form can be saved printed or exported to Excel 101 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved de Visualyzer 1 1 Default Style Internal KrackHighTec_M eng DDAR Fie Edit Layout Query Analysis Reasoning Help 1203082 SC BOM 299 HEAD MORZA Th BH HRS y IRelations ADVICE Total nodes 21 Ena
18. cannot coexist with a change in the attribute for even one node via the Display Settings form Should you attempt to effect a change a warning to this effect will be received 70 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ee Visual yzer 1 1 Default Style Internal Tree_with EdgeAttributes Types _M eng DAR File Edit Layout Query Analysis Reasoning Help enbas 01 BOM 229 LAA DRA E Attribute based Label Color Shape Size Link Color amp Daniel Node attribute Se James MARRY A Sarah Legend Abigail DB divorced E married E single David os be P Alyssa a a Yen E Joseph i Les l E e MARRY GENDER E divorced O Female 4 ee BB married O male Matthew BB single Nodes 18 Links 17 Use Dbl click to add or edit a nodejlink right button add linkfaccess command menu left button drag select disable node Figure 58 Legend window Views A view consists of the last set of attribute queries which can be saved and later applied to the same or other data This allows you for instances to retain a uniform look and feel for the results of many interviews conducted in the course of a study The command to save or to load is under the Query Views menu Fig 59 figure below It is recommended that views be named according to the effect they produce 71 VisuaLyzer
19. data 121 130 211 231 311 321 21 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved If the data represents friendship nominations then the first line indicates that node 1 Paul nominates node 2 Liz as a friend This above input data will generate the corresponding adjacency matrix in VizuaLyzer Note both binary and valued relations can be represented using the edgelistl format A single edgelistl file can represent multiple relations each of which are separated by a vertical bar or exclamation sign Acquiring edgelistl data from UCINET is a two step process exporting the files from UCINET and importing them into Visualyzer Exporting from UCINET 1 Select Data gt Export DL in UCINET This will open up the data export window as shown below Export as DL file Input dataset ADVI CE Output format EDGELIST1 Diagonal present Present edgelist only Type Directed Decimal places 9 Field width Freefield Guarantee space YES Page width E 0000 Embed row labels INO Embed column labels NO Embed matrix labels INO Dutput dataset ADVICE txt Figure 14 Exporting from UCINET 2 Select input dataset and most importantly chose edgelist format from the output format drop down box 3 The data type drop down box lets you select whether the data is directed asymmetric or undirected symmetric 4 Give th
20. di 0 In this case the was_infected_by relation can be partitioned into two subsets a deterministic as expected part was_infected_by complement was_infected_by di and a nondeterministic unexpected part was_infected_by was_infected_by di These constraints can partition the network and even discover missing knowledge The menu choices are Determinism and Nondeterminisms from within the Reasoning menu 109 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Consider a very small network involving nodes b e fand g The following was_infected_by relation is not completely deterministic so this network does not model an incurable disease The network is automatically divisible into two parts a deterministic part in black and a nondeterministic part in red Visual yzer 1 1 WIB eng S Oj xj File Edit Layout Query Analysis Reasoning Help 1 MidB HAL S Hod OLS PSSM YO M was_infected_by M PRIOR deterministic was_infec BB PRIOR nondeterministiciwas_ir gt Rename Apply Colors Select All Unselect All Link attribute Select attribute value s _ Select All Unselect Close Figure 82 An infection network for a contagious curable disease Nodes 4 Links 8 Use Dbl click to add or e The solution can be either a general relation or within the specific network depending on the users choice Relation complem
21. field on the Size screen that is displayed Link Color To have link color be based upon the value of one of the links attributes e select Query gt Attribute Based gt Link Color e or click the Link color by Attribute icon This will display the Link Color tab of the Attribute Based screen Select the attribute you want used to determine link color in the drop down list box at the top of this screen The Legend box will then be filled with a list of the values for that attribute with a colored square next to each value indicating the color that will be used for links with that value Fig 57 69 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Attribute based TIT Label Color Shape Size Link Color Link attribute Co_workers Legend A false H true C Gray scale Figure 57 Attribute based selection screen link color tab You can change the color used for an attribute value by double clicking on the colored square to the left of it and then clicking on a color from the Select Color screen that is displayed Legend for attribute based queries The legend window displays the last attribute queries for node label color shape and size and link color Figure 58 The legend can be saved as a bitmap or JPG image and used later in conjunction with the graph image in presentations Query based attribute settings
22. help define other primitives The apartness relation di relative zero or diversity 1s complement id Absolute unit is id di and zero 0 is id di The relative sum R amp S is complement complement R complement S Finally an explicitly defined set is a brace delimited comma separated collection of pairs that are each colon separated such as fann carl carl bill In an infinite universe the complement of that finite set would be infinite However complement fann carl carl bill has a finite description as a collection of equations and inequations in disjunctive normal form x yl x ann y carl V x carl y bill V x ann x carl This constructive complement satisfies the usual properties of Boolean algebra and relation algebra If a set description uses an existentially quantified variable in an equation its complement would require a universal quantified variable in the corresponding inequations For example complement p 3 1 NAx NAyl for_every NAI NA2 NAx p NAy NA2 NA1 where in such a set description for_every NA1 NA2 means that NAJ and NA2 are universally for all quantified As expected of a classical complement it is particularly true that complement complement p 3 1 p 3 1 Variables in VisuaLyzer are a hybrid of the traditional notion of a value that is Not Available NA and the mathematical symbols x and y A variable name is chosen by the following rules 1 it consists of the two charact
23. indicates indirect knowledge through an intermediary Any such unknown third person is a candidate for introduction But there is more to be said The potential introducers or sponsors of such unknown candidates are those intermediaries They are related by R complement R converse R which is understood to mean those that I know but who know someone that I don t know the can_be_sponsored_by relation It is sometimes useful to identify the extremities of a network In a hierarchy these may be the leaves of the tree or the root Every nonempty relation has sources and sinks sometimes the relation has sources that are not sinks and may have sinks that are not sources These more general notions of extremity include hierarchies as a special case and appear at the network fringe They are also easily defined as respectively is_linked_to complement converse 1 is_linked_to and is_linked_to complement converse is_linked_to 1 In a network of transference of respected opinions the sources may be pure followers and the sinks may be pure leaders That is a pure follower should not lead but should follow at least one leader in dual fashion a pure leader should not follow but should lead at least one follower The position leaders are those pure leaders who are not followers in a network derived by ignoring collegiate links that is relative duals The network of Figure 77 has four individuals in a totally dependent subnet but who
24. of the AND gate current can flow only if both switches A and B are closed The expression A B represents the corresponding product intersection of sets In the circuit made of the OR gate current can flow if at least one of A or B is closed The expression A B represents the corresponding sum union of sets Similarly the conjunction of two sentences A AND B is true only if both sentences A and B are each true the disjunction of two sentences A OR B is true only if A is true or if B is true or they are both true As should be apparent there are isomorphisms between logic set theory and digital circuit Although AND and OR are fundamental they are not enough as primitives for Boolean logic The additional operation NOT negates any value to its opposite The following truth table shows the outputs of NOT from its inputs 104 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved A NOT A 0 1 1 0 As one would expect the NOT Boolean function maps sentences that are TRUE or FALSE into corresponding sentences that are FALSE or TRUE Similarly the set complement maps empty and universal sets into universal and empty sets The notation for the complement function on a set S is complement S It is straightforward to extend these concepts to sets of more than one element These larger sets correspond to an extension of the bit operations to more tha
25. of the network and attributes of individual nodes For example 93 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved it is not surprising that the CEO actor 7 and one of his Vice Presidents actor 14 is in their own block because of their unique relations with other managers who are in the other major block 3 5 As with other procedures the user can cycle through and select each of these positions for closer examination using the Select position box on the form 6 The results from the roles and positions form can be saved printed or exported to Excel 7 The form provides two other options critical to Roles and Position analysis density and image matrix ese Visual yzer 1 1 Default Style Internal KrackHighTec eng DER File Edit Layout Query Analysis Reasoning Help ee a PR SS Select g gt ES H a letwork Properties PE re Eh E Image Node Centrality matrix Adjacency Matrix Nearest Neighbors Shortest Path E bel EN Cliques Relations ADVICE Partitions Clusters Partitions levels 21 Partition level 3 Roles and Positions PANA Note Link directions ignore Y Y k qn dd dd Opinion Leaders PINUS E we fil Position 1 19 1 A Select 12 Density 13 is matrix 16 17 18 19 2 20 21 3 4 Is Core and Periphery P Position 2 1 7 Position 3 1 14 Number of
26. parameters you specify To create a new random graph select New Random from the File menu A dialog box will appear see Fig 2 above Specify the number of nodes you want in your graph in the Number of Nodes field Then specify either the number of links you want in your graph in the Number of Links field or the field When one of these link fields is changed the other will automatically adjust itself When you have your node and link parameters set click the OK button to see your graph An example of a random graph is shown on page 10 Fig 3 Creating Nodes and Links VisuaLyzer also allows you to create nodes and links on the screen interactively To create a node simply double click anywhere on the screen To create a link between two nodes right click on the first node hold the right mouse button down drag the link to the second node and then release the mouse button see Fig 6 above The direction of the link will be from the first node to the second node You can also link a node to itself Node Attributes VisuaLyzer allows you to change the names of your nodes and to add attributes to nodes To edit a node s information right click on the node and then select Edit Node from the pop up menu This will bring up the Edit Node screen Fig 10 Double clicking on a node will also bring up the Edit Node screen To change the name of the node type the new name in the Name field at the top of this screen The node type can also b
27. researchers with a number of network analysis functions Along with some basic network parameter estimation researchers can also calculate cliques partitions communities shortest paths nearest neighbors and roles and positions Researchers can also submit queries to locate nodes and links that fulfill specific criteria For additional analysis researchers can export their data to UCINET EXCEL GraphML and other formats VisuaLyzer has been developed by Medical Decision Logic Inc with support from the National Institute of Drug Abuse NIDA through the Small Business Innovation Research SBIR Phase II project A Tool for Network Research on HIV Among Drug Users R44 DA012306 System Requirements Disk Space Approximately 18 MB of disk space is required to install VisuaLyzer including XSB programs Operating System VisuaLyzer is designed to work with Windows 2000 and Windows XP its behavior on other operating systems may be slightly different Screen resolution Screen resolution should not be less than 1024 x 768 pixels While not required we recommend using a screen resolution of 1280 x 1024 or higher XSB logic programming environment In order to perform some of the logical path analyses XSB has to be installed on your local machine The XSB package is included in the VisuaLyzer setup program No additional installation is necessary Installation After downloading running the file VisuaLyzer2 0setup exe will install
28. screen Selection of a style will only affect links with a pixel width of 1 Font You can change the font of the labels of the selected link s by clicking on the letters or blank space to the right of the Font label in the Links section at the bottom of the Nodes Links tab of the Display Properties screen This will display a font selection screen Figure 44 Select the options you want for the font and then click the OK button to change the link label s Arrowhead Color You can change the color of the arrowheads of the selected link s by clicking on the colored square to the right of the Arrow Head Color label in the Links section at the bottom of the Nodes Links tab of the Display Properties screen and then selecting a color from the Select Color screen that is displayed This is similar to selecting colors for other options previously described Head Length You can change the length of the arrow heads of the selected link s by entering a number representing a number of pixels in the field to the right of the Head Length pixels label in the Links section at the bottom of the Nodes Links tab of the Display Properties screen and then pressing the Enter key Head Width You can change the width of the arrow heads of the selected link s by entering a number representing a number of pixels in the field to the right of the Head Width pixels label in the Links section at the bottom of the Nodes Links tab of the Display Properties sc
29. the nearest neighbors of a node in VisuaLyzer follow these steps 1 Open a graph and select Nearest Neighbors from the Analysis menu to open the nearest neighbor form as in Figure 71 x Visual yzer 1 1 Default Style Internal KrackHighTec eng uery Analysis Reasoning Help Neighbors SuN B y Node 20 Relations ADVICE Neighbors of 20 Groups found 2 p a i a a a a a a a a a a a ab ab oe kek a ao Group 1 17 Distance 1 1 Had eS wt ei Select group 1 fe Nodes 21 Links 190 18 nodes selected 144 links selected Figure 71 Nearest neighbors 2 Select a node to find its nearest neighbors e g node 20 in the Krackhardt advice network shown above 3 VisuaLyzer returns two groups of nearest neighbors Group 17 nodes are within 1 geodesic of node 20 In other words node 20 and any of these 17 nodes are separated by one geodesic while the 20 nodes in Group 2 are two geodesics away from node 20 86 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Shortest Paths A geodesic is a shortest path There may be more than one shortest path or geodesics connecting any two nodes Shortest paths indicate the possible paths that network contents information infections ideas practices etc can pass efficiently from one node to another and the extent to which this process can be constrained by the nodes th
30. want to import either double click on it or single click on it for it to be displayed in the File name box and then click the Open button Your data will be displayed as a network graph Importing data into an existing file If you have already created a VisuaLyzer file and wish to update the information without recreating your layout and display settings you can do so by importing new nodes and or attributes This means that if you are conducting an ongoing study you can import analyze and manipulate your data and network graph at any point and then update it periodically with the latest information Adding nodes by merging graphs To merge two graphs first open the graph to which new nodes will be added An example named Tree with EdgeAttributes txt from Examples UCINET folder is shown in Figure 20 30 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved a James Sarah gt A a Daniel Next select Add from the File menu and select the file with the new nodes This file can be in any format accepted by VisuaLyzer In this example Tree with EdgeAttributes Add txt was selected from Examples UCINET folder John Q Joseph O Alyssa Christopher e Ochi Emily nn Lauren annah oc AD Alexis Matthew Y Nicholas Samantha A David b Elizabeth Figure 20 Tree with edge attributes 31 VisuaLyzer User Manual v 2 0 Copy
31. 13 txt Square grid 5x5 txt E Square grid 6x6 txt B Symmetric_Diamond txt E Symmetric_Triangle txt E Tree 90x90 txt El gt Filename 3 z My Network Fies of type UCINET edgelist1 edgearray1 fles Figure 15 Importing from UCINET Select UCINET edgelist1 edgearray1 files in the Files of type box to have these files displayed When you have found the file you want to import either double click or single click on it for it to be displayed in the File name box and then click the Open button Your data will be loaded and displayed as a network graph If the graph contains any self loops VisuaLyzer will prompt you to ignore the self loops or to keep them Note that to import networks with different link types you can use File gt Add function several times Or you can use File gt Import to import the first network and then File gt Add for the others GraphML files GraphML is a comprehensive and easy to use file format for graphs a variant of XML adapted to describe graphs It has predefined words to mark graph s properties such as node edge source target directed etc and its files can be read it by any XML compatible web browser e g IE Notepad and can be processed by XML parsers The format is convenient for exchanging data with other visualization tools or MDL software such as QBuilder Further information about GraphML its history and background can be found at http g
32. 7143 stror Rename Apply Colors Note Link directions ignored Dokdo ok ok Asymmetries Select All Unselect All Core nodes not found Completions Link attribute Unique Links Tests Select attribute value s Arbitrary Relations X Cancel Operation Unselect All Nodes 7 Links 6 The links at the core of the network Figure 86 The network middle is in black although there is no core A relation is said to be symmetric if it is equal to its converse Thus a test for symmetry is simply a test of equality between the two relations If the relation is infinite it must have a finite representation Ina sociomatrix of a finite network this means that the upper diagonal is a reflection of the lower diagonal This is an intuitive description but does not apply to infinite relations A relation is Symmetrized up by summing the relation with its converse R converse R A relation is Symmetrized down by forming a product of the relation and its converse R converse R Similarly we can find that asymmetric part that is needed to Symmetrize up by subtracting the complement from the converse converse R complement R or find the extra asymmetric part by subtracting the complement of the converse R complement converse R If the relation is a network the graph completion is the set of all possible pairs of nodes of the original network Thus a network involving n nodes has square of n number
33. Boolean Dual of R R di as R complement R di Peircean duals don t change true to false but instead change left to 113 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved right The followers leaders example above and the endpoints and start end links examples illustrate a Peircean duality The core periphery procedure described in the previous section sometimes cannot find a core or periphery set of nodes This is compared to the procedure for finding the fringe subnet and the remaining middle subnet The fringe subnet is simply the sum of the start edges and the end edges The middle subnet is the rest of the network that s not the fringe This comparison is illustrated in Figure 86 Those links between the nodes 1 2 and 3 comprise the network middle although the network has no core or periphery e ee Visual yzer 1 1 Default Style patientstatus wiring bd File Edit Layout Query Analysis Reasoning Help SABHAL d ems PS l aM HBL A BO dh Select Relation Links Nondeterminisms gt Capa i 5 Introductions gt Ml is_linked_to a 3 A S wh E PRIOR middle is_linked_to Posfions vee Rabon 6 red Start End Links gt Total nodes 7 Ensblsdacdaei_ 7 Fringe Middle Links Find the network Fringe start_links R end_links R 0 0 000 7 100 000 Self loops Find the network ani iii m ue Symmetries Core centralization 0
34. C oO oO So OV Oe m 1S eo oom o rO000rrOorO0000000 0000000000000 0000000000000050 500000000000000000 ODO F OOGO DDD 0 OD 000 rFrO0O0000000000 OFPFFO0O0O00000000000 Figure 26 Adjacency matrix screen There are several options for saving your matrix The matrix will be saved as you see it on the screen You can still edit the matrix while this window is open Click the 5 button to update the matrix in the window to match the changes you made to the graph Click the H button to save the entire matrix Click the Click the 4 button to save only the upper right half of the matrix Vv button to save only the lower left half of the matrix 0 Click the E button to fill the matrix diagonal with 0 s 1 1 Click the button to fill the matrix diagonal with 1 s Click the button to fill the matrix diagonal with s 37 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Click the a button to toggle saving the node labels Click the bi button to bring up the standard Windows save screen Type the name of the file to be saved in the File name field and click the Save button If you wish to export your adjacency matrix to Excel set up the file as you want it and then click the export to Excel button This will open Excel and paste in the adjacency matrix You can then name and save the file in Excel Figure 27 shows an adjacency matrix for a simple 2 mo
35. Layos iii 11 Sprins embedding TOUS A AA A E 11 A Asay OU tog lt cadedas Ye densa taunts atsacadasptesh da vee San ba cmiecnaiasuaad sane neh Sosnaeena N re a a Ea 12 Adding INOS e a E a A A E a a E a T 12 Adding LINKS arrn eieren a REA A REE T A E E AE E A EAT R 13 MONA a OS TOR E Masts oan A E O 14 Display PrOpErUES aan SS A E E EER ER 14 Enterimg Data into Visuals y ZeP wicca scscscensessscevdacsdssvcdsnnatr cesses sudiecssavddeudeussesusvcs odcseiuadssseuundssccensssaveds setiva sasso reese 17 Creatine Random ALA sccy accel odes e ts oa saa aeons add a tas es 17 Creatine Nodes and as a 17 Node ADUE cao SOU UR E E a 17 Link Types and AUTIDUES odds 19 Multiple LINKS ui didas 20 impor DEN 21 UCINE T dgelistl edscarray e a 21 Graph ME files iii aida 24 A sacs sttnsiedacks cegndacequansainisecnat venous aghe a gray e EAEE AS ESTE EEOAE 26 Microsott CS Td Sac hs a a bisa en 27 CS V lles commiaseparated Values csi o O ici 29 O TO 30 Importing data into an existing lei A adi secci 30 Adding des DY merging A RR a E E a RA sours 30 Adding new Node Attribut S creii tai A A E A te eave edi E ER E es 34 Data AAA NI sdasheir sdebdussisosatussacesetissdisaliestsacdessedsddas 35 AO A a ed o E a OS 35 DAVIN CAG O RN 36 EXPO A A A Aes tment 36 UCINET Edgelist los cito idos dto 36 UCINET Edecarrayl ascii td a dirias 36 A o al a he va Sade Srcads 36 Dy Neth 0 A A ee ay hati 36 Microsoft Excel a 36 Prolog esca e e a e i ta ar re a a i a eS a 37
36. To add node attributes text files should be in TAB delimited format using quotes when spaces are present 34 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Data Export There are several options for saving or exporting your data each of which is described below Saving files To save your data as a VisuaLyzer file which uses a eng extension select Save from the File menu If you had previously saved the file the file will be saved If this is the first time you ve saved the file this will bring up a standard Windows Save screen Fig 25 Save Graph Save in Examples Binary csv My Recent DyNetmL Documents Excel GraphM Reasoning O Sequence S UCINET My Network Save as type Visual yzer file teng Figure 25 Saving data Type the name of the file to be saved in the File name field and click the Save button To save a file with a different name select Save as from the File menu to bring up the Windows Save screen to enter the new file name 35 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Saving images To save your graph as an image select Save image from the File menu This will bring up a standard Windows Save screen Type the name of the file to be saved in the File name field select the image type you want to save to in the Save as type dro
37. User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ee Visual yzer 1 1 Default Style Internal Tree_with EdgeAttributes Types_M eng DAR File Edit Layout Query Analysis Reasoning Help en Attribute Based P yo ES 2 Ha 40 HDN EAA Legend Ctrl L Load and Apply Vie Cte Attribute based Q Search Query Ctrl Q Save Current View N Label Color Shape Size Link Color Linked Pairs Comparison a Node attribute story 5 pre Daniel AGE x James Legend O Size Value 0 10 12 14 20 16 23 17 24 4 David 18 26 21 32 22 33 da _ Nicholas 24 37 em amantha 2 e al r Alyssa E John Joseph ad MARRY GENDER O E divorced O female E married O male ji 7 j Alexis H single Matthew A bal Cese Nodes 18 Links 17 Sizing based on the AGE attribute Figure 59 Loading and saving views Display Relations Types Since VisuaLyzer allows multiple link types relations you can choose to show any number of these relations as well as the color of each of them To control which relations are shown and their colors select Select Relation from the View menu or click the Select Relation button This will display a Relation tab on Select Relation Type screen Fig 60 12 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights r
38. VisuaLyzer 2 0 Default Style Internal Status EdgeTypes_M eng File Edit Layout View Query Analysis Reasoning Help ien ABMS ACR BROOME SPSL uano HWHR2z2Z HOaD Cutpoints 1 node cutsets 5 Compute a a a Relations is_linked_to m Sharon Total nodes 14 Enabled nodes 14 Total components 1 Isolates a Dyads 0 Susan Components 3 1 Note Link directions ignored Add dd dto aa a a a Cutpoints found 2 Cutpoint Total Isolates Dyads Comp3 Abs Disintegration Rel Disintegr Charles 4 3 D 1 23 08 4 00 Stuart 2 0 0 2 7 69 2 00 AA Tanya Doma Es Wynn Select cutpoints Nodes 14 Links 24 2 nodes selected Figure 77 Cutpoints References e S Wasserman and K Faust Social network analysis Methods and Applications Cambridge University Press Cambridge 1994 96 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Opinion Leaders Over the past 50 years theories on diffusion of innovations have been used to study the spread of new ideas information and practices in a variety of applied settings Opinion leaders have been recognized as critical elements in the diffusion process largely because of their ability to reach and diffuse ideas information practices to their followers Among the techniques used to select opinion leaders sociometric techniques have assumed a central place given their utilization of interpersonal comm
39. VisuaLyzer 2 0 User Manual o se ae Ye ole gt o awe e gt Lo a eo N in Le Ne o 7 oOo E pon AA o o Ne e A oe o o LN SocioMetrica VisuaLyzer has been developed by Medical Decision Logic Inc with support from the National Institute of Drug Abuse NIDA through the Small Business Innovation Research SBIR Phase II project A Tool for Network Research on HIV Among Drug Users R44 DA012306 Copyright O 2005 2007 Medical Decision Logic Inc All rights reserved February 15 2007 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved TABLE OF CONTENTS TABLE OE CONTENTS ai ad ERA arado pecador 2 TABLE OF FIGURES iaa A US iS Ee Ea aeea LI S 5 Medical Decision Logic MC cesconarsucesuesvsscanscopapedesasdocsnessareee peiores ossos oso oneen tasso iost 7 S cioMetrica VISUALLY ZE id 8 System Requirements seidsessssees seseneccuentvevncesiaceccosucussesntavexnosdvavedsnacevavesiencesutave csenansbesssnaenceonsuesdssosaoycaossveeed sessies ee 8 DISPO ld a S 8 A OS 8 A nadas 8 ASB logic programming ENVITONMEN diarias dol 8 AA AA nn nn A 8 Registration A eonccwanedanees tune deca aesuaneneveuau ousudeassuuess suwucelactovauas cece cdencesevsudsvereuanecsvonideraeaes 8 GUIDED TOUR OF VISUAL Y ZERiteccscisceitiscstentesoncetasscucasgnsedsctsecnscea casdcastoscaeceedesous eiaceontencnoccesisaeseaadeebnssedasesn 9 Croatia a CA A A as S aTe 9 Using Graph
40. VisuaLyzer on your system Registration Optional registration will insure notification of future developments VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved GUIDED TOUR OF VISUALYZER The following pages will give a brief tour of the VisuaLyzer program and describe many of the features that are available More detailed information about specific functions and processes can be found later in the manual You can access help for any VisuaLyzer screen or function by pressing the F1 key or selecting Help from the Help menu Please note that the screens you see on your screen may look slightly different than the screens shown in the manual due to differences in operating systems and display settings Also note that throughout this User Guide we will use the term graph to refer to the display of the nodes and links in a network Creating a Graph When you first run the VisuaLyzer program you will see the screen below de Visual yzer 2 0 Default Style Default Untitled DAR File Edit Layout Yiew Query Analysis Reasoning Help ISAR BRS AJP BOOMS 22 HEAD NORA REGAR Nodes 0 Links O Use Dbl click to add or edit a nodejlink right button add linkfaccess menu left button drag select disable nodes Figure 1 First screen The fastest way to create a new graph from scratch is to select New Random from the File menu 9 VisuaLyzer User Manual v 2 0
41. ads ada deine 12 Fig re GO Adin a HOG eo ta ed ll ed e de 13 Figure 7 AQUEL A A A A A A 14 Fig re Changing display properties A dd 15 Bisure 9 Changed display Properties sintio 16 Figure 10 Addition of a new node attribute cinc db dial 18 Figure 11 Toggle node attributes window Ss toi id 19 Figure T2 Edit link STE A A A E a A Ae A 20 Figure 13 Multiple links between nodes iii a 21 Figure 14 Exporting from UCINE T dl a t i 22 Fig re13 importan tron UCI PTS EAEE AS aaa 24 Fig re 16 Importn from EXE y n A ie N tos 27 Figure 17 Importing adjacency matrix from Ex e e e ee ide eo 28 Figure 18 Importing from CSV Mes ies iin ana adinen ds 29 Figure 19 Opening Saved VisuaLyser files A a A Gl ee ae Reads 30 Figure 20 Tree with ede e attr DUtes nenene Ohya A A A Beaty a ede 31 Figure 21 Addition of nodes to existing graph cccscccscesssesstetoserscencesseteresscesnsesssctacessaccessasessnaccsensacessaesenss 32 Higute 22 Square ond 5X3 graphml cese atada accio Cut es perdia a dl des Aig eds ca kal dee edo 33 Figure 23 31 Cube PAPA 5 cise ok vesetias dae hee ded A a AS 33 Figure 24 Addition of existing nodes only adds new liMkS oooonnnccnoccnooncnonononnnononononcnononnnnnnnn conocio non ncnnnnnnncrnn anno 34 Figure 25 Saving dal to de 35 Figure 26 Adjacency matrix Screen sii ai 37 Figure 27 Adjacency matrix for movie by actor affiliation network seseseessesesesessesresseseresressesnresrensessrerreesesse 38 Figure 23 Movin
42. al tab in the Display Properties selection screen should be un checked 66 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Attribute based Label Color Shape Size Link Color Node attribute AGE v Legend Miz G20 Das Mos H 26 LJ Ms 33 37 Ml ss A 1 Gallapse values Ed WES 110 Select value range Start value End value Color 12 Select Color Figure 54 Selecting group of values color tab Node Shape To have node shape be based upon the value of one of the nodes attributes select e Query gt Attribute Based gt Shape e orclick the Attribute Based Query button and select the Shape tab This will display the Shape tab of the Attribute Based screen shown below Select the attribute you want used to determine shape in the drop down list box at the top of this screen The Legend box will then be filled with a list of the values for that attribute with a shape next to each value indicating the shape that will be used for nodes with that value Fig 55 67 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Attribute based ps Label Color Shape Size Link Color Node attribute GENDER Legend O female male Figure 55 Attribute based selection screen shape tab
43. and preloaded into VisuaLyzer by the file userdefined pl as described in the next section The figure shows four layers of the NCI ontological hierarchy hierarchy cdfclasses 1 CDF Classes CAD i tiii i 1 1 1D f 4 BicChChHDiagnostic_aEO FindindFindings_and GerPro Anatomic_StructurClinical_or_Research_Activity Gana GburgicaRetired_Clinical_or_F5SuGene_wMajor_ProtReelTRetired_OlRetired_Org firus Concepts Figure 87 A portion of the NCI ontological hierarchy For the hierarchy called generalizes the subsequent steps include menu selections 115 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved e Reasoning gt e Hierarchies gt e Show hierarchically generalizes Additionally if there are several hierarchies each is displayed after Show and within the parentheses after hierarchically So one of several hierarchies may be displayed from the list The preceding figure depicts four layers of a portion of the NCI ontology defined as a generalizes relation Limitations and further information Note that VisuaLyzer does not display node attribute values that have attributes even though selectors in relative products can create such deep nesting of attributes Additionally relation selectors in VisuaLyzer extract attributes and can create some nodes with attribute names different from others If nodes have different attribute name
44. at lie on those paths The procedure implemented in VisuaLyzer gives the number of all shortest paths geodesics connecting given pair of nodes and provides a way to view all of them one by one To find the shortest paths between any pair of nodes in a graph follow these steps 1 Open a graph 2 Select Shortest Path from the Analysis menu This will open up the Shortest Path form Fig 72 for the Krackhardt advice network 3 Select a given pair of nodes for example nodes 16 and 14 4 VisuaLyzer will indicate the number of possible shortest paths found between the two nodes in this case 6 and their length in this case 2 Each of the six possible paths is listed in the main window of the form and the user can use the Select Path control to zoom in to each of the six possible paths and see their trails in the graph Results can be printed or saved in text or excel file formats for further processing ele Visual yzer 1 1 Default Style Internal KrackHighTec eng DAR File Edit Layout Query Analysis Reasoning Help SAHE Newokroperies Pe aR Mo WEN 2 EAR Node Centrality Adjacency Matrix Nearest Neighbors Shortest Path ly p Shortest Path undirected graph x S HRS From 16 Cliques Partitions Clusters gt Roles and Positions ini To 14 Opinion Leaders L Core and Periphery x Relations ADVICE 7 t A 14 Shortest paths found 6 Length 2 Note Li
45. ate an ideal core periphery structure This measure of fitness is similar to the Pearson Correlation Coefficient UCINET uses to judge the adequacy of the observed core periphery pattern with the idealized pattern e The current version of VisuaLyzer uses a simple matching coefficient for matrices A and B instead to reduce the computation time The coefficient is expressed as r SUMiSUM f Aij Bij where the function f Aij Bij returns 1 0 when Aij Bij Aij lt gt Bij for both core and periphery regions of the matrix If the region s data are totally opposite to the model then r 0 if they are identical r 1 e Normalization only includes the sum of nodes in the regions Diagonals self loops and disabled nodes are excluded from the analysis e The Core and Periphery procedure produces two lists of nodes for core and periphery and normalized correlation coefficient r named core centralization that shows how well the real structure approximates the corresponding model 100 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved The core centralization coefficient r is constrained to the range between 1 to 1 to be comparable with the measurements based on Pearson s coefficient Cases with r gt 0 9 are marked as perfect fit cases r lt 0 1 and lt 0 are marked as no fit To detect a core periphery structure in a network follow these steps 1 Open a network graph
46. b for k n every individual has been reassigned to a different block and G n f0 where fO is the fitness at the beginning of the current Step 4 loop Step 5 Choose the value of k for which G k is maximuml If G k gt 0 the local neighborhood solution is given by the partition obtained from the initial partition by the definite reassignment of al K ak and Steps 1 through 5 are repeated resetting the locks If G k lt 0 then we are done 102 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved References e Borgatti S P Everett M G 1999 Models of Core Periphery structures Social Networks 21 375 395 http www analytictech com borgatti cp_structure doc e John P Boyd William J Fitzgerald and Robert J Beck Computing Core Periphery Structures and Permutation Tests for Social Relations Data September 28 2004 Institute for Mathematical Behavioral Sciences Paper 16 http hypatia ss uci edu imbs tr mbs04_13 pdf 103 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved FORMAL REASONING ABOUT SOCIAL NETWORK DATA Links as Facts The viewpoint taken in this section is a little different Many approaches to network analysis take as fundamental either an adjacency matrix or an edgelist Additionally the network editing portion of VisuaLyzer assumes that the network drawn is the founding knowl
47. bedding Layout from the Layout menu press the F9 key or click the Spring Embedded Layout button below the main menu Fig 3 This will re organize the graph Fig 4 which was created with 14 nodes and 16 links de Visual yzer 2 0 Default Style Default Untitled File Edit Layout View Query Analysis Reasoning Help SAd HHS A Ow BeOS 2294440 HERAA ThE Node_006 Click the circular layout E Node_002 icon NS si a Se _AQNode_012 J SQMode 0075 aA N Node_o01 a O Node_011 j 3 Quede 005 Node_010 O Mode_004 j Node_003 j A Node_013 Node_008 E Node _009 Nodes 14 Links 16 Use Dbl click to add or edit a node link right button add linkfaccess menu left button drag select disable node s Figure 4 Spring embedded layout Note that the disconnected node Node_006 is shown in the upper left corner of the screen All disconnected nodes are by default displayed in the upper left corner of the screen so that they can readily be found and moved into the network if desired The spring embedding process may take a minute or more for dense or large graphs but can be stopped by clicking the Spring Embedded Layout button or pressing the F9 key again 11 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Circular Layout Next we will organize our graph using a circular layout To do this select Circular Layout from the Layout
48. bled nodes 21 Core nodes 13 61 905 Periphery nodes 8 38 095 Core Peripheri ratio 1 6250 Core centralization 0 7547 sl Note Link directions ignored paias i d aa a a a a a ator att ka ok ea ak a Core nodes 13 1 Periphery nodes 8 gt 2 Show O Core nodes Periphery nodes Click to select only peripheral or core nodes Nodes 21 Links 190 Figure 80 Core Periphery analysis Kernighan Lin Algorithm for the Core Periphery Problem The gain function g a represents the fitness gain from moving a one of the n individuals from its current block core or periphery to the other block The outer loop Step 5 is typically repeated only two or three times Step 1 Pick a random core periphery bipartition with at least individuals in each block Step 2 Choose an individual a such that g a is maximum even if not positive Step 3 Perform a tentative reassignment of a to the other block Step 4 Repeat Steps 1 and 2 exactly n times where an individual cannot be chosen to be reassigned the locking rule if he she has already been reassigned in one of the previous iterations of Steps 1 and 2 but within the current loop of Step 4 This sequence of reassignments defines a sequence of gains g1 K gn where gi corresponds to the reassignment of individual ai to the other block in the ith iteration The total gain after k reassignments equals G gt gt g1 N
49. bove the query selects pairs of nodes that are more than 20 years apart in age Click Submit to execute the query This will present the results in a graph like the one shown below 78 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Linked Pairs Lomparison Select link if the Following conditions applied to the differen nodes that connected by the link Example AGE gt 10 Select ink the difference bel AGE attribute values of endpoint nodes exceeds 10 yd Attribute Function Diff Value Diff value2 Age gt 20 i Figure 66 Results of linked pairs comparison The results of the linked pairs comparison can be represented as a binary similarity matrix for further analysis processing The cells in the matrix take the value of 1 if the age difference between pairs of nodes is greater than 20 years Query History The Query History window presents a list of previous queries these can be edited and or resubmitted for analysis 79 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ANALYZING SOCIAL NETWORK DATA Network Level Properties Visualyzer provides a number of network properties by default To see these properties follow these steps 1 Open any network graph eng file E g Krackhardt s HighTech Advice network graph 2 Select Network Propert
50. c relations and have little value alone The dichotomizers apply in combination with some other relation For example for a network R the product R di would find the subset of R without self loops Similarly the constants would likely combine with other relations as part of more complex expressions Currently any relational expression can be entered by performing Reasoning gt Arbitrary Relations gt Execute Relational Expression Ctrl R and typing the expression into the box Please note the hint at the bottom of the screen displayed in Figure 73 in which the result is displayed Doing so pops up a text box into which the expression is entered In particular to find who reports to a manager but doesn t seek his advice we calculate the value of the expression reports_to complement seeks_advice_of The resulting discovery is that manager 11 reports to manager 8 but manager does not seek the advice of manager 8 that is the pair 11 18 is a member the only member of the combination reports_to complement seeks_advice_of In the reports_to network there is a link between and 78 but there is no link between 77 and 8 in the seeks_advice_of network The only stated fact linking and 8 is that 11 reports_to 18 In VisuaLyzer the conclusion is the black link in Figure 81 107 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ae VisuaLyzer 1 1 Krackhard
51. ces among actors VisuaLyzer calculates closeness centrality for only undirected or symmetric graphs The sum of these geodesic distances for each actor is the farness of the actor from all others The reciprocal of farness that is one divided by the farness gives a measure of nearness or closeness centrality This is normalized as a percent to the maximum possible nearness For a 7 node network 1 Star network the center node has closeness 100 and the six peripheral nodes 54 55 2 Circle network all nodes have the same closeness 50 3 Line network the two end nodes are less close 28 6 than those in the middle 50 84 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Node Ce pay sh was A Betweenness Centrality for undirected graph Node Betweenness Normalized Betweenness 9 074 4 7 76 6 489 3 415 6 410 3 37 4 3 101 2 666 2 522 2 274 1 851 1 827 1 728 1 617 1 556 0 722 0 659 0 568 0 439 0 348 0 290 0 171 0 160 0 149 1 629 1 289 0 149 4 776 Maximum possible betweenness C Auto update Figure 70 Node properties betweenness Node Betweenness Centrality Betweenness centrality indicates the extent to which an actor falls on the geodesic paths between other non adjacent pairs of actors in the network Betweenness centrality views an actor as wielding power over interactions between othe
52. complement converse reports_to 1 6 Who is at the top is the manager in chief converse reports_to complement reports_to 1 1 7 Are there any friendships A yes 1 or no 0 question I is_friends_with 1 8 Who are those individuals with more than one Friend is_friends_with is_friends_with di 9 Who are those individuals with only one Friend is_friends_with complement is_friends_with di 10 Who supervises Someone who considers no one a friend Who has unfriendly Subordinates E g 14 9 converse reports_to complement is_friends_with 1 To understand query 10 consider that Who supervises Someone is equivalent to Someone reports to Who m since the supervises relation is basically converse reports_to The relative product is_friends_with 1 says that Someone is friends with Everyone and complement is_friends_with 1 states that Someone is friends with no one Therefore reports_to complement is_friends_with 1 gives us the unfriendly Subordinates of Supervisors The converse of the previous set is the set of Supervisors of unfriendly Subordinates Solving for Unknowns in Deterministic Networks A directed network is deterministic if each node has an outward edge to at most one node That is the maximum out degree of the network is one In the previous section query 9 finds the deterministic subrelation of the is_friends_with relation and query 8 finds the nondeterministic subrelation It is often valuable to kn
53. connected The graph is directed Current relations ADVICE Degree Centrality Node Degree InDegree OutDegree Degree InDegree OutDegree normalized 18 15 17 160 000 75 000 85 000 15 11 130 000 75 000 55 000 4 20 120 000 20 000 100 000 9 14 115 000 45 000 70 000 18 105 000 90 000 15 000 13 105 000 65 000 40 000 3 100 000 40 000 60 000 5 100 000 25 000 75 000 3 100 000 40 000 60 000 5 100 000 25 000 75 000 13 95 000 65 000 30 000 10 90 000 50 000 40 000 4 85 000 20 000 65 000 4 75 000 20 000 55 000 11 70 000 55 000 15 000 10 70 000 50 000 20 000 9 70 000 45 000 25 000 3 60 000 40 000 20 000 10 55 000 50 000 5 000 4 50 000 20 000 30 000 7 45 000 35 000 10 000 90 476 45 238 45 238 27 685 19 849 26 613 9 1 45 000 20 000 5 000 32 160 000 90 000 100 000 e C Auto update Figure 68 Node properties degree centrality Degree The degree of node is the total number of edges incident to the node A symmetric or undirected graph simply has a nodal degree value as we cannot or do not want to distinguish in degree from out degree The degree is important because it tells us how 82 VisuaLyzer User Manual v 2 0 Out Degree In Degree Normalized Degree Copyright 2005 2007 Medical Decision Logic Inc All rights reserved many connections an actor has For example actor 18 in the graph has the most advice relations with others
54. cular division is meaningful the algorithm defines quality function or modularity Q as follows Let Eij be the fraction of edges in the network that connect nodes in group i to those in group j and let Ai Sum Ey Then Q Sum Eii Ai 2 Thus the modularity is the fraction of the edges in the network that connect nodes within community minus the expected value of the same quantity in a network with the same community divisions but random connections between the nodes If the number of within community edges is no better than random we will get Q 0 Values approaching Q 1 which is the maximum indicate strong community structure In practice values for such networks typically fall in the range from about 0 3 to 0 7 Higher values are rare The algorithm falls in the general category of agglomerative clustering methods It assumes that graphs are undirected with unweighted edges Starting with a state in which each node is the sole member of one of N 89 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved clusters we repeatedly join clusters together in pairs choosing at each step the join that results in the greatest increase or smallest decrease in quality function or modularity Q The worst case running time of the algorithm is O M N N or O N22 on a sparse graph where M is number of links and N is number of nodes in the network The algorithm has the added advantage of ca
55. d from the calculations Size of components created by the removal of a cutpoint varies from just one node to the half of the graph To estimate the degree of graph fragmentation after cutpoint removal we use two measures absolute and relative disintegrations Assuming there is 100 disintegration if we get all N nodes of the graph disconnected absolute disintegration will be TotalComp 1 N 1 If the component count is 1 there is no fragmentation Relative disintegration is defined as a ratio of total component numbers after before cutpoint removal We leave the final decision about an importance of each particular cutpoint to the user 95 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved To run the Roles and Positions procedure follow these steps 1 Open a graph a Note Any single graph with unweighted edges is supported in the current version of VisuaLyzer Future versions will support multiple and valued relations b VisuaLyzer will symmetrize the graph automatically using the maximum procedure if the graph is directed and will transform it into dissimilarities before submitting it to the cutpoint procedure 2 Select Analysis gt Cutpoints from the menu This will return the positional graph with selected cutpoint nodes if any are found two nodes have been identified in Fig 77 3 The results from the Cutpoints form can be saved printed or exported to Excel m
56. de network showing a movie by actor relations Adjacency Matrix SJ Compute bel a Y NODE TYPES movie actor Simple N x M matrix N 3 Shining The Batman A Clockwork Orange COLUMN HEADERS l Jack Nicholson 2 Stanley Kubrick Figure 27 Adjacency matrix for movie by actor affiliation network You can save print this matrix or export to excel following the same procedures as described above for the single mode network Display Options Selecting nodes There are several ways to select nodes and links in VisuaLyzer The simplest way to select a node or link is to click on 1t If you hold down the Shift key while clicking on nodes and links all you clicks will be selected together You can also select several nodes and links by drawing a box around them To do this click on a blank part of the graph and hold the mouse button down This creates one corner of your box at the spot where you clicked Drag the opposite corner of your box to somewhere else on the screen All links and nodes completely contained within this box will be selected 38 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved To select all nodes you can e use Ctrl A e or click Select All button on the toolbar e oruse Select All in the Edit menu Note that whenever two nodes are selected and they have a link between them the link will also be selected Graph Layout Visual y
57. dit a nodejlink right button add link access menu left button drag select disable node s Figure 43 Node labels used as node shapes Note that when node images are displayed they will be displayed on top of the node shape except when the label is the shape 55 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Image You can change the image displayed for the selected nodes s by clicking on the image or blank box to the right of the Image label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen There are 15 predefined images to choose from or you may load a bitmap or JPEG image file of your choice To make the selected image transparent thus showing the node shape or background behind it check the Transp checkbox below the word Image To remove the image completely click the blue word Clear below the Transp Checkbox Note that when node images are displayed they will be displayed on top of the node shape except when the label is the shape Size You can change the size of the selected nodes s by typing a number representing a number of pixels in the field to the right of the Size pixels label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen and then pressing the Enter key You can also change the size of the selected node s by using the up and down arrow keys on your keyboard Label Orien
58. e changed and is set to actor by default To create a new attribute for a node click the New button A new row will be created in the attribute table on the screen Type the name of the new attribute in the Name cell and type the value of the attribute in the Value cell For example in Figure 9 node 006 s age is 32 When a node attribute is created it is assigned to every node in the graph but with a value of na or not available until the user changes it You can edit an attribute s Name or Value by clicking on the cell and typing in the new information You can delete a node attribute by selecting that attribute s row and then clicking the Delete button When you delete a node attribute you are given two choices e you can either delete the attribute from all nodes in the graph e or change the value of the existing node s attribute to na and not delete the attribute and values from all the other nodes in the graph soft delete 17 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ele Edit Node Name Node_006 Type actor New User defined Attributes Attribute Age Figure 10 Addition of a new node attribute You can link the node with a web page or document for instance person s resume photo image or spreadsheet data Just type the path to the destination in the value field like in the following exa
59. e Disable helps to explore affiliations in the network and conduct What if types of analysis To control which node types are shown and their colors select Select Type of node tab on the Select Relation Type screen from the View menu or click the Select type of node button This will display the Type of node tab on Select Relation Type screen Fig 61 73 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Select Relation Type Relation type of ink Type of node E movie Ml actor Unselect All Action Enable Select Node attribute Select attribute value s Select All Unselect All Figure 61 Select type of node The lower part of the screen refers to node attributes and works as described above for the Relation type of link tab However the action choices on this tab are enable select rather than show hide Graph Animation VisuaLyzer allows you to animate your networks by displaying several eng files sequentially similar to a cartoon flipbook All files involved in the animation sequence must be located in the same directory VisuaLyzer will display all files in the chosen directory in alphabetical order To animate a series of graphs select Animate Sequence from the File menu This will display the Network Animator screen shown below 74 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision L
60. e File name field select DyNetML files in the Save as type drop down list box and click the Save button Microsoft Excel To export your data in Excel format select Export from the File menu to display the standard Windows Save screen Type the name of the file to be saved in the File name field select Microsoft Excel Workbook in the Save as type drop down list box and click the Save button 36 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Prolog files To export your data in Prolog P format select Export from the File menu to display the standard Windows Save screen Type the name of the file to be saved in the File name field select Prolog files in the Save as type drop down list box and click the Save button Adjacency Matrix To export your data as an Adjacency Matrix select Adjacency Matrix from the Analysis menu An option for a Regular single mode network and one for an Affiliation 2 mode network will pop up Figure 26 shown below refers to a single mode network Adjacency Matrix 013 Pad a be ty Qu Ha tn n co w w e b h Susan Kathy Tanya Donna Nancy Manuel Harold Charles Carol Wynn Bob Fred Sharon Stuart 03M ns 0nNr 0000000000000 DA DE N A EA A O VA DA E A oO rr O O O OO O OOO O O 00 ODO OOCL 0c 0c ODO Ee BS OOo Oo So eG ee aaa ee eee rFPoooccec crc 0c OF
61. e R o a ac coh cats 91 Fig re 70 Roles and POSITIONS A A A R seat A eate 94 Figute 77 Cutpoms 220 nusndmved A A a ies 96 Figure 78 Sel ctimg numberof CAMCES 225 5 tsei e itise eani loa 97 Figure 79 Opinion leaders selected DDIES ii A A AA A A E ds 98 Preure su Core Perphery abla ySis a te cad A a A EE ve ae dea SETE lieben cates 102 Figure 81 Determine who reports to but does not seek the advice of which Manager ocooconnccnnncnnoninonnnannnnnnos 108 Figure 82 An infection network for a contagious curable disease ooooooncccninociconocononoccnoncnononcnonnnnnonnncncnnncnnnnnnno 110 Figure 83 Finding the unknowns NA within a deterministic subnetwork ooooconccnnoccnococoncnonononnnononononccnncnnnnoo 111 Figure 84 Solutions in red for missing victims NA Of b oooooccnocccoccconanononononcnnnnnonn ccoo nono nonnnnnrnnnrnn cnn nernn cnn 112 Figure 85 Links to Position leaders are in black Opinion leaders are in red coooooccnnnccionononcnananonanononcnoncnanc conan 113 Figure 86 The network middle is in black although there is NO COLE eee eeeeeeeeeeseeceseceseeeeeeeeaeecaeeneesseeeeaees 114 Figure 87 A portion of the NCI ontological hierarchy 000 0 ee ceseeeeeeescecsseceseeeseeesseecaeceseesseeesseecsaecseesseeeenees 115 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Medical Decision Logic Inc www mdlogix com Founded in 1997 Medical Decision Logic MDLogix is a growin
62. e Show ToolTips check box on the General tab of the Display Properties screen is checked a node s or link s attributes and values will be displayed as a yellow tool tip when the cursor is held over them When this box is not checked this information will not be displayed This affects both all the nodes links currently in the graph and all new node links Show Links When the Show Links checkbox on the General tab of the Display Properties screen is checked all links will be displayed When this box is not checked links will not be displayed This affects both all the links currently in the graph and all new links Show Link Arrow When the Show Link Arrows checkbox on the General tab of the Display Properties screen is checked all directed links will have an arrowhead indicating the To Node of the link When this box is not checked no links will have arrowheads This affects both all the links currently in the graph and all new links 51 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Show Link Type relation When the Show Link Type checkbox on the General tab of the Display Properties screen is checked all links will be labeled with their type as defined on the Edit Link screen When this box is not checked these types will not be displayed This affects both all the links currently in the graph and all new links Show Link Label When the Show Link Label checkbox on t
63. e is a Partitions clusters form which lists membership for each cluster 90 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved 3 The Number of clusters option on the form can be used to instruct VisuaLyzer to select a particular number of clusters in the graph For example in Figure 74 VisuaLyzer displays 3 clusters for the advice relation graph 4 Additionally the Select cluster can be used to select all members of a particular cluster Upon selection the user can apply different attributes color size shape to all cluster members to distinguish them from the other clusters as the example above shows In addition a cluster number is printed at the edge of each cluster to further distinguish the clusters 5 Contents of the Partition Clusters form can be saved printed or exported to Excel for further optional processing Finding Communities As stated above this algorithm divides a network graph in to various communities or components that may be in existence in the natural community The fit of the implementation is examined through the modularity Q index the higher the value the closer the divisions approximate those occurring in the natural community efe Visual yzer 1 1 Default Style Internal KrackHighTec eng File Edit Layout Query Analysis Reasoning Help COBRAR rr O Pp Aa HN a LAA Node Centrality Adjacency Matrix Nearest Neighbors S
64. e node images revealing the shapes behind them by un checking and checking the Show Images checkbox in the middle of the General tab of the Display Properties screen This affects both all the nodes currently in the graph and all new nodes Note that when node images are displayed they will be displayed on top of the node shape Note that if you uncheck both Show Shapes and Show Images that the shapes will still be displayed in order to ensure that the node is visible Show Labels Checking and un checking the Show Labels checkbox on the General tab of the Display Properties screen will show or hide the labels of the nodes This affects both all the nodes currently in the graph and all new nodes The Toggle Off On button on the tool bar can also be used to turn the labels off and on Show Borders Checking and un checking the Show Borders checkbox on the General tab of the Display Properties screen will show or hide a thin black border line around each node This affects both all the nodes currently in the graph and all new nodes Show Disabled Nodes When a node or link is disabled described below it is displayed as light gray To completely hide disabled nodes un check the Show Disabled Nodes checkbox on the General tab of the Display Properties screen Checking this box will again cause disabled nodes and links to be displayed in light gray This affects both all the nodes currently in the graph and all new nodes Show ToolTips When th
65. e of weight 1 the higher is the rank of the nominee and such a nominee will be listed first as a member of the group e Note If all links to nominees have the same weight or are even unweighted as in the example above then group members are selected based on previous criteria only point 5 above 7 The output of the Opinion Leaders procedure is the list of leaders and list of group members for each leader together with additional parameters showing selection reason and sorting order References e Thomas W Valente Rebecca L Davis Accelerating the Diffusion of Innovations Using Opinion Leaders ANNALS AAPSS 566 November 1999 e Thomas W Valente Social Network Thresholds in the Diffusion of Innovations Social Networks 18 69 79 e Thomas W Valente Robert K Foreman Integration and Radiality Measuring The Extent Of an Individual s Connectedness And Reachability in a Network Social Networks 20 1998 89 105 e Thomas W Valente cctrial_kv4 prg a GAUSS program that finds opinion leaders and assigns them leaders to groups 99 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Core and Periphery Core periphery analysis is central in network analysis because it offers interesting insights into social stratification Who are the core actors of the network Who the peripheral Intuitively core nodes are connected to each other more or in many more ways than with
66. e output file a meaningful name under output dataset option Remember edgelists1 files are text files and therefore must have the txt extension 5 Click the OK button to save the file Note the location of the file so you can copy it into your VisuaLyzer data directory 6 The header and fragments of the exported edgelist file looks like this 22 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved DL N 21 FORMAT EDGELIST1 DIAGONAL ABSENT DATA 121 141 181 1161 1181 1211 261 271 2211 2121 2131 2141 2161 2171 2181 21 121 21141 21171 21181 21201 7 The header entries provide information about the dataset a DL indicates it is a DL or UCINET file format b N 21 indicates that there are total of 21 nodes in the matrix C d e FORMAT EDGELIST1 DIAGONAL ABSENT indicates that the DL file is exported using the edgelistl format with diagonal entries of the graphs suppressed absent DATA marks the beginning of the actual data Exclamation or vertical mark at the end of the file indicates the end of one relational matrix separating it from other relational matrices in the case of multi relational datasets 8 This procedure will export either a single multiple relation data Exporting multi relational datasets follows the same procedure except that the resulting edgelistl text file has to be edited slightly to accommodate difference
67. e the color of the selected nodes s by clicking on the square to the right of the Color label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen This will open a Select Color screen Click on a colored square and the selected node s will change to that color Shape You can change the shape of the selected nodes s by clicking on the shape or blank box to the right of the Shape label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen and then clicking on a shape from the Select Shape screen that is displayed Fig 42 54 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Select Shape X Figure 42 Select shape screen Clicking on the A on the Select Shape screen assuming that all nodes are selected will cause the node s label to be used as the shape this means that the label will be centered between any links to those nodes as shown below te Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng File Edit Layout View Query Analysis Reasoning Help enb WAS Daniel James N En _Sara Abigail f David X i a mE Elizabeth Samantha eo Micholas A X Alyssa John 2 hs E n Hannah Emily a Christopher ji A ri Joseph i j E Lauren f Y Alexis Matthew Nodes 18 Links 17 Use Dbl click to add or e
68. e top of the General tab of the Display Properties screen This color is changed like the background color is changed click on the box to open the Select Color screen and then click on a colored square to select that color Note that changing this color does not affect the color of existing nodes only new ones Group Node Color The color of the node representing newly grouped nodes see group node section can be changed by changing the color of the box next to the Group Node Color label at the top of the General tab of the Display Properties screen This color is changed like the background color is changed click on the box to open the Select Color screen and then click on a colored square to select that color Note that changing this color does not affect the color of existing grouped nodes Selection Frame Color When a node is selected it is surrounded by a colored square and when a link is selected it changes to this same color You can change the color indicating that node or link is selected by changing the color of the box next to the Selection Frame Color label at the top of the General tab of the Display Properties screen This color is changed like the background color is changed click on the box to open the Select Color screen and then click on a colored square to select that color This affects both all the nodes currently in the graph and all new nodes Cluster Frame Color The color of the frame outline of node cluste
69. edge In this section sentences as facts are fundamental and networks or matrices are derived from the stated facts at least conceptually From these facts we can set up equations on relations and solve for network unknowns using the axioms of a relation algebra Statements of fact are simple sentences of the form Subject verb Object Predicates verbs relate a source Subject with a sink Object Directed edges in a network can also depict these sentences The predicate is the relation that names a set of directed edges The nodes are the subjects and objects Such simple sentences are assertions of what is known and true a missing or unasserted sentence is assumed to be false Then the model is explicitly available for criticism as to how complete or consistent it is The logical operations do not directly apply to the real world Instead they apply very strongly and very precisely to this model of the real world Boolean operations Boolean algebras are more general than relation algebras because there are fewer values and fewer operations involved In the simple Boolean logical systems there are only TRUE and FALSE values In the binary systems at the foundations of digital computing there are only O and 1 bits Also there are three basic operations Boolean functions AND OR and NOT Truth tables give precise meanings for these operations In a digital circuit O may represent an open switch and 1 a closed switch In the circuit made
70. ee 0 are excluded from the calculation of the ratio 4 To find out the M opinion leaders the program compares the in degree of the persons If in degrees are the same it further compares the relative integration Valente 1998 of the candidates The integration criterion shows the degree to which an individual s inward nominations integrate him her into the network The integration is a centrality measure that is based on the so called inverse Guimares Relative Integration Index or reverse distance matrix calculations cited in Valente 1998 The M leaders then are matched to the N M persons in the community who nominated them 5 We assume the size of group for each leader is approximately the same N M M If someone did not directly nominate a leader that person is matched to a leader who is closest by distance to him her If he is equally close to two or more leaders then indirect paths are used to determine optimal pairing geodesic distance 2 During data processing the procedure ignores repeated nominations i e when a person nominated other person more than once it also ignores self nominations 6 If weights or values are assigned to links for example when users are asked to list important nodes first followed by less important ones the first nominee is assigned a value of 1 the next 2 and the last N In such a case selection of group members will reflect the nomination ranks e Note The lower the weight e g the cas
71. electing AGE instead of GENDER will result in the display in Figure 31 ee Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng File Edit Layout View Query Analysis Reasoning Help 1208 AGS ACH BOOMS SPAS LEM DRAA HOD Abigail X Chris apher Daniel Joseph Six Tau an Lauren Nicholas David pavigeth ox ies John Samanth ass E Emily S Jacob Nodes 18 Links 17 Use Dbl click to add or edit a node link right button add linkfaccess menu left button drag select disable node s Figure 31 Layered Layout graph 2 Now it is easy to see which nodes are oldest and youngest and which are older and younger than others but it is difficult to see how they are connected because most are in the same position along the X axis We can fix this by going back to the Layered Layout setup screen and changing the X axis layout We can set the X axis to minimize edge crossings by selecting the based on optimization algorithm option in the X axis box Fig 32 For the Y axis you can organize your nodes also according to node degree number of adjacent links as well Together with the descending Sort Order option this will show nodes with more links at the top of the screen and nodes with fewer links at the bottom This kind of layout makes it easy to see which nodes are the most connected 42 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision
72. ent is classical and taken over any possible term So the most general solution to for example Who was_infected_by b could even be some unnamed individual not in the original group and the solution would indicate this through the inequality constraints On the other hand if we want to limit our attention to only the current network then relation complement is too general and a simple graph complement will suffice Consider a specific network with one unidentified node labeled as Not Available NA As described earlier VisuaLyzer Reasoning considers NA s to be variables and NA values can be discovered if a global network constraint such as determinism can be imposed The most general solutions should include individuals that even may be outside the network and those solutions can be found with relational descriptions in menu choices such as Reasoning gt Determinisms gt Find the deterministic subrelation R complement R di On the other hand it is more reasonable to limit these unknowns to the local community described in the original network The following is the converse of a sample infection network that assumes the disease is incurable 110 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved i xelx File Edit Layout Query Analysis Reasoning Help gt gt gt ki Find the deterministic subrelation R complement R di Nondeterminisms Find the deter
73. enu This will bring up the Display Properties screen shown below 48 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Display Properties SS General _Nodes j Links Background Color Background Image Clear Node Color Group Node Color E a Selection Frame Color B Cluster Frame Color Position Frame Color E Show Background Image Show Node Shapes C Show Node Images Show Node Labels Show Node Borders Show Disabled Nodes Show ToolTips Show Links Show Link Arrow C Show Link Type relation C Show Link Label C Show Link Weight C AutoOrganize Animation C Sound Effects Figure 39 Display properties screen Figure 39 shows the General tab on this screen the Nodes Links tab will be described later Also note that this window can remain open on top and moved as you work with the Visual yzer display For example you can create and drag nodes while this window is open You can save your current display settings by clicking the Save button and you can load previously saved settings by clicking the Load button and then choosing the setting file from the standard Windows open file screen When you load display settings they are applied to all nodes links and the background If at any time you wish to return to the original display click on the Reset Display Settings button on the toolbar If only the last change is not desired use the Undo butto
74. er Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Query Node Attributes Link Attributes Set node state Enabled Disabled Selected if the Following conditions applied ad nadas Attribute Function value Yalue2 N a 8 1 Age 40 AND Level 2 PA 4 Nodes 18 and 21 satisfy the query condition they are the only 2 managers under 40 years old in the Krackhardt s advice network Figure 64 Result of node search The above example finds nodes based on attributes If the graph has link attributes e g frequency of communication you can run queries that find nodes that satisfy a specific condition e g nodes that communicate daily or weekly In this case you select the Link Attributes radio button The right hand panel lists the nodes or links according to their state selected enabled or disabled The Remove button will remove the selected conditions whereas the Clear All button deletes all of them This 1s particularly useful when queries have to be quickly modified and run again Linked Pairs comparison Linked pairs comparison allows a researcher to select nodes based on the difference in their respective attribute values For example a researcher may be interested in age differences among married couples as a correlate of marital stability To explore the data the researcher may want to find all couples who are at least 10 years apart in age The
75. ers NA in any case mixtures or 2 it begins with NA in any mixture of case and is followed by any number of characters the first of which is chosen from 0 9 x y z Both node names and attribute values may be NA Not Available If nodes have no attributes then their names represent them If nodes have attributes then they have a deeper term structure of the form node name NameValue attributel Value1 attributeN ValueN The relation operations form a highly symmetric arrangement and comprise the operations of a structure described in Tarski amp Givant as a representable relation algebra RRA 106 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Combinators Dichotomizers Directors Involvement Products Reciprocals Constants Absolute R S complement R xi yi Relative R S i i converse R pm The relation operators divide into three major groups of four and each group is orthogonally divided into absolute or relative object involvement Combinators combine or integrate together the knowledge from two different relations The Dichotomizers are useful to control separation in a set of pairs For example the absolute unit relates everything with everything while the relative zero di diversity or apartness relates only those objects that differ The final group of Directors either reciprocate or fix the arguments In particular
76. es screen is checked certain actions such as automatic layouts will be accompanied by sound effects When this box is not checked no functions will be accompanied by sounds Node options VisuaLyzer provides several options for displaying nodes These options can be accessed and set by e Pressing the Display Properties button e Pressing the F4 key e Selecting Edit gt Display Properties gt View Edit from the main menu This will bring up the Display Properties screen for Nodes Links Figure 40 52 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Display Properties SESE General Nodes Links Nodes Selected 0 Color Shape Image E Transp Clear Size pixels Label Orientation Font Links Color Width pixels Style Font Arrow Head Color Head Length pixels l a ly Head width pixels tal El Load Settings Save Settings Close Figure 40 Display properties Nodes Links Note that we are looking at the Nodes Links tab on this screen the General tab was described earlier This screen shows the information for the currently selected node s No nodes were selected when the above screen shot was taken thus there are no values are displayed No value information is also displayed when multiple nodes are selected and have conflicting values e g they are different colors or shapes When a node is selec
77. eserved Select Relation Type Relation type of link Type of node lis linked to C Ml go_shopping_with O M have_kids_same_school Link attribute Select attribute value s Figure 60 Select relation links The top part of the screen indicates which relations are being displayed and the color of the links for each Check or un check the checkbox next to a relation to show or hide that relation link in the network graph For example in Figure 60 three link types are shown The lower part of the screen refers to link attributes In the case being considered the links have no attributes Suppose however that the graph showed only one relation is linked to had been assigned an attribute of friend with values of True or False Friend would have been listed in the Link attribute drop down box and the choices of True False and na in the Select Attribute Value s panel Clicking the check box next to true would have shown links to all the people who were friends You can change the color used for an attribute value by double clicking on the colored square to the left of it clicking on a color from the Select Color screen that is displayed and then clicking the Apply Colors button Similarly since VisuaLyzer allows multiple node types actors events etc you can choose to enable or select nodes of specified types Selecting nodes is useful for setting visual and custom attributes for a group of nodes Enabl
78. eses enclosed comma separated list of terms of the form f t1 t2 tn Terms themselves have a representation as a rooted directed ordered acyclic network With relations come concepts of query and computation We can define new relations as expressions combining other relations and operators Not only can we test for membership in a relation but we can query for the members as pairs or provide one object and query for the other The underlying model is a collection of sets of ordered pairs of terms a collection of networks of networks and unifies high level software with the underlying low level hardware circuitry Relation operations The knowledge base consisting of a collection of facts and displayed as directed networks is a sub theory of the theory of relations Relations are sets of ordered pairs and operations on them satisfy the rules of classical set theory specifically both Boolean and relation algebras The basic operations on relations are 1 absolute sum or union of relations R and S is R S fx ylx R yor x S y 2 absolute product or intersection of relations R and S is R S x yl x R y and x S y 3 relative product composition of relations R and S is R S x y there exists z x R z and z S y 4 the complement of the relation R those pairs not in R is complement R x y x R y is not true 105 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved
79. esults Because of the traditional finite binary representation of floating point numbers presumably true equations may not hold One example is 10 3 3 10 Note that an undirected network is considered a symmetric directed network for these link reasoning purposes An undirected link between two nodes is an assertion of both a forward direct fact and its converse Thus a sentence such as ann and bill play cards together would better be described as the pair ann plays cards with bill and bill plays cards with ann The installation of VisuaLyzer creates an empty file in the default fundamentals folder called userdefined P The user can define other relations such as a large logical database in this file and it will be loaded at startup An example of a logical database of two facts is the file userdefined P containing the following op 950 xfy learned_of_the_innovation_from ann learned_of_the_innovation_from bill bill learned_of_the_innovation_from carl 116 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved The userdefined P file is loaded very quickly On a Windows XP Pentium 4 system a network of 100 000 edges can load in only a few seconds Finally the menus of VisuaLyzer show the ampersand amp character as lt ampersand gt For example the relative sum is not shown as R amp S but is instead shown as R lt ampersand gt S Any relational expressions entered involvi
80. etwork The graph shows clique 1 with green colored diamond shaped nodes 88 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ee Visual yzer 1 1 Default Style Internal KrackHighTec eng File Edit Layout Query Analysis Reasoning Help 17 H 52 Network Properties 4 Node Centrality adjacency Matrix Clique Configuration Panel Nearest Neighbors Shortest Path Cliques a A N 7 Pl A BS am Sy EN H Clique min size 5 El Roles and Positions a __ e i Relations ADVICE Opinion Leaders gt Clique min size 5 Partitions Clusters Cliques found 43 Note Link directions ignored iodo a a e aa a a a a Clique 1 9 1 Core and Periphery 15 118 2 20 21 Clique 2 10 1 10 11 15 l18 12 20 3 14 la 3 Select clique 1 fal Nodes 21 Links 190 Use Dbl click to add or edit a nodeflink right button add link access command menu left button drag select disable node Figure 73 Cliques Clusters and Communities VisuaLyzer implements the Newman and Girvan algorithm for finding clusters communities in graphs see references below The algorithm is based on the idea of modularity Given any network the clustering algorithm always produces some division of the nodes into communities regardless of whether the network has any natural such division To test whether a parti
81. f the nodes will be double that of the others Note that the Display Properties window will always stay on top of the graph The General tab lets you change display properties for the whole graph For example you can display link weights by clicking the Show Link Weight checkbox or turn on off animation of nodes 15 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved EOR VisuaLyzer User Manual v 2 0 fe Visual yzer 2 0 Default Style Default Untitled File Edit Layout Yiew Query Analysis Reasoning Help rA Sp BOOMs LPS REMY WHEN A4 BGK 1223 ARMS O Node_006 lade_005 ee we lt n pi NA 004 A Y Ne 003 Ghede o 002 Ed pu TN f E lode_001 Node_008 1 A Node_007 Node _009 a da A LA Nee 010 E a MM spel Noe oo 2 nodes selected Figure 9 Changed display properties Nodes 15 Links 17 Node_015 and Node_001 have been doubled in size and link weights are displayed 16 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Entering Data into VisuaLyzer There are several ways to enter or import data into VisuaLyzer You can create links and nodes directly have the VisuaLyzer create random data for you or import data from several popular formats Each of these is described below Creating Random Data VisuaLyzer allows you to create a pseudo random network graph based on
82. g INKS A E EE A A AR 39 Figure 29 Layered Layout setup screen ii ib is 40 Fig r 30 Layered Lavoro 41 Eigure 31 Layered Layoutoraph asl i a e od 42 Figure 32 Layered avout Setup Screen Lic ra osas 43 Figure 35 Layered Tay OUTDOOR A A RA A daa 43 Figure 34 N mode Network Layout Options yisc sccsccessaccesassdeessetescectascececstivavatausudvecsaueaceasisoeeaccashasseva sedseuaateccacctens 44 Fig re 35 Simple bapa tite gr ph cacao A nda sea ol Sle ede ce 45 Eli ure 30 Gb egos Mi E ERLE Ne aA ETON ud 46 Figure 37 Best fit for cube Sraph ini cs ia dd Asaltan aras dedican 47 Figure 38 Selected nodes position Al ti iia 48 Figure 39 Display properties SCleen v 22 itis A A agate 49 Figure 40 Display properties Nodes Links A a 53 Figure 41 Node properties on display Crec tontas td 54 Figure 42 Select shape SCreen sissies dai ta dl 55 Figure 43 Node labels used as node SHapes a ida 55 Figure 44 Font selection Creen ii A E 56 Figure 45 Enable nodes Screen ssiri et needed et atl ae Rea aed ernie 57 Pisure 46 Hat group Sr tdi 58 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Figure 47 Edit proup screen members 42 742 h esc ee A Betas ic ER ce eee ud 59 Figure 48 Display properties screen links ii ias is 60 Figure 49 Edit attributes for group of links ii ai 62 Figure 50 Basic Pr A ld 63 Figure 51 Adding meaning to a basic graph A eA eS Ae eR 64 Figure 52 Attribute base selection EA ia
83. g information technology company that develops software for public and clinical health markets We create software architectures components and systems to optimize information processes in health research and practice Our mission is to use information technology e To apply public health science knowledge and methods in administrative clinical community and occupational setting and e To link patients clinicians researchers administrators and others to enhance health and productivity Our goal is to serve customers and researchers by providing Effective systems for risk detection and preventive interventions Cutting edge logic technology for diagnosis and decision models Products designed to increase efficiency and quality in clinical settings Automated generation of billing codes and required documentation VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved SocioMetrica VisuaLyzer SocioMetrica VisuaLyzer 2 0 subsequently referred to as VisuaLyzer is designed to graphically display small and mid sized social networks Researchers can import their data from UCINET edgelist or edgearray GraphML and other formats into graphic network of nodes and the links connecting them Once displayed the visual properties such as color shape size and location of the nodes and links can be customized to create an informative graphic representation of the data VisuaLyzer also provides
84. g up this screen 19 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ste Edit Link Relation is_linked_to o New From To Node Node_O001 To From Node Node_o02 Directed Oves no Label Link_001 Weight 11 000 User defined Attributes Attribute Web Www google com Figure 12 Edit link screen This screen displays the information about the selected link Link attributes function just like node attributes for adding editing and deleting them On this screen you can also change the link s Label and Weight by changing the information in those fields You can indicate if a link is directed or not directed by selecting the corresponding Directed radio button You can change the To Node and From Node of a link by selecting these nodes in the corresponding drop down list box To assign a web page or document to a link type the path to the destination in the value field Clicking the Open button will open the page or document using the default viewer program Fig 12 Multiplex Links VisuaLyzer also allows multiple link types or relations This means that one node can be connected to another node multiple times with each link being of a different type If you create a link between two nodes that are already connected the above Edit Link screen will be displayed If you want your new link to be of a different type click the label New under the Re
85. hardt s high tech example A researcher interested in career trajectories of young executives may want to find the young e g under 40 yrs managers in the company s advice network To execute this query load the Krackhardt s advice network and select Search Query from the Query menu This will bring up a window like the one shown below 1 Select the basis of the query node or link attributes 2 Select node state whether you want the nodes enabled disabled or selected Query Node Attributes Link Attributes Set node state Enabled Disabled Selected if the Following conditions applied Selected nodes Attribute Function Value Yalue2 A crap g ristopher BOE 5 40 Daniel AND Level 2 Hannah Joseph AND Sarah AND Remove Figure 63 Node search 3 Set the conditions e Select an attribute of interest from the drop down box under the attribute column Select a mathematical function e g less than greater than equal to etc Select a value for the attribute You can perform Boolean queries based on multiple attributes In the example above the query selects nodes that are younger than 40 years old AND who are at level 2 managers of the corporate hierarchy e Click Submit to execute the query This will present the results in a graph like the one shown below 76 VisuaLyzer Us
86. he General tab of the Display Properties screen is checked all links will be labeled with their Label property as defined on the Edit Link screen When this box is not checked these labels will not be displayed This affects both all the links currently in the graph and all new links Show Link Weight When the Show Link Weight checkbox on the General tab of the Display Properties screen is checked all links will be labeled with their Weight property as defined on the Edit Link screen When this box is not checked these values will not be displayed This affects both all the links currently in the graph and all new links AutoOrganize When the AutoOrganize checkbox on the General tab of the Display Properties screen is checked every time a node is moved or a link is added the graph will automatically organize itself according to the last selected automatic layout spring embedded circular or layered When this box is not checked nodes and links will stay where they are placed Animation When the Animation checkbox on the General tab of the Display Properties screen is checked each time a layout is applied including via the AutoOrganize feature both the nodes and the links will be shown moving from their current positions to their final positions When this box is not checked nodes and links will simply snap to their final positions Sounds Effects When the Sound Effects checkbox on the General tab of the Display Properti
87. hese two operations find possible values from NA limited to those already in the network The determinism constraint applies to unknown link sources not destinations In this case the solution for the missing victims of b that is fc d e is the red network and the remainder of the network is the black network 111 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Visual yzer 1 1 WIBreslts eng File Edit Layout Query Analysis Help Select Relation Links xj eoe i gt eal gt i Es Ea xl A Ml was_infected_by WB PRIOR nondeterministiciwas_infected_ BB PRIOR deterministiciwas_infected_by gt Rename Apply Colors Select All Unselect All Link attribute Select attribute value s te Select All Unselect an Close gs Saved Nodes 16 Links 37 Figure 84 Solutions in red for missing victims NA of b Often a network satisfies or should satisfy a global constraint that can be described as such a relational expression That expression can then separate the network into two parts and along the way possibly solve for missing NA values Illustrative Examples Example number 3 of the previous list of examples has an associated query Example 3 found candidates for introduction as the relation R R complement R di If the relation R indicates some form of directed personal knowledge then R R
88. hortest Path 3 B iy Cliques Relations REPORTS_TO Partitions Clusters Hierarchical Clustering pear were 2 Roles and Positions Optimal Partition Communities Note Link directions ignored N Stoo kk ik ko a a ook Opinion Leaders Community 1 13 1 Core and Periphery 10 11 12 16 17 18 1 Community 2 8 13 14 15 19 20 Select l community 1 Nodes 21 Links 20 Use Dbl click to add or edit a nodeflink right button add linkfaccess command menu left button drag select disable node Figure 75 Communities 91 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved To divide a graph into communities 1 2 Open a graph Select Analysis gt Partitions clusters gt Optimal Partition Communities menu as it is shown in Figure 75 The Communities form displays the name of the relation number of communities in the graph a message box will indicate if no communities are found the modularity index Q as well as options for the user to cycle through and select each of the community members for closer examination or editing change color size and shape The example above implements the algorithm on Krackhart s REPORTS_TO relation a component of the High Tech Managers study A closer examination of the data gives a sense of natural divisions based on departmental affiliation tenure and to some e
89. ies from the Analysis menu to open the Network Properties form Fig 67 ete Visual yzer 1 1 Default Style Internal KrackHighTec eng File Edit Layout Query Analysis Reasoning Help 22 gt h E E Peano WPN Node Centrality Adjacency Matrix Nearest Neighbors Shortest Path Cliques Partitions Clusters Roles and Positions Opinion Leaders Core and Periphery Network Properties Suma y The graph is connected The graph is directed Total nodes Enabled nodes Isolates 0 Dyads 0 Components 3 21 Groups O Relations Current relations ADVICE Total links 312 Current links 190 Current enabled links 190 Diameter 2 Average geodesic distance 1 3095 Density 0 9048 Degree Centrality 10 526 Closeness Centrality 48 018 Betweenness Centrality 3 304 all measures for undirected graph Node attributes summary numeric categorical AGE 39 714 DEPT 2 190 LEVEL 2 714 TENURE 11 746 STD Count 27 000 0 000 1 000 0 250 value 9 326 1 139 0 547 7 846 lt Auto update ADVICE FRIENDSHIP REI MIN Propor 62 001 4 000 3 000 30 001 Nodes 21 Links 190 Total nodes Enabled Nodes Use Dbl click to add or edit a nodeflink right button add link access command menu left button drag select disable node Figure 67 Network Properties Total number of nodes in current network s For example there advice network and all n
90. in degree to the number of actors in the network less one expressed as a percentage Actor 10 for example has outgoing ties with 70 of the remaining actors This is a figure we can compare across networks of different sizes Unlike degree centrality which only takes into account direct connectivity to immediate ties closeness and betweenness centrality takes indirect ties into account Measures for these are produced and explained below 83 VisuaLyzer User Manual v 2 0 Node Closeness Centrality Node Centrality 3 x HN Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Closeness Centrality For undirected graph Node Farness Closeness 15 20 0 2 21 0 18 23 0 3 120 AVG 26 190 STD 3 486 IMIN 20 0 MAR 33 0 Maximum possible closeness IES 0 050 0 048 0 043 0 043 0 043 0 042 0 042 0 042 0 040 0 040 Normalized Closeness 100 000 95 238 86 957 86 957 86 957 83 333 83 333 83 333 80 000 80 000 76 923 76 923 76 923 74 074 68 966 68 966 68 966 66 667 64 516 62 500 60 606 77721 10 306 60 606 100 000 0 050 C Auto update Figure 69 Node properties closeness Closeness centrality is based on geodesic distance and measures how close an actor is to all other actors in the network An actor is central if it can quickly interact with all others One could consider either directed or undirected geodesic distan
91. is will return a new graph with the socio metrically derived opinion leaders selected Fig 79 97 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Visual yzer 1 1 Default Style Internal OpinionLeaders eng DER Fie Edit Layout Query Analysis Reasoning Help J 2ARBARS 0l EROM PPSF atmo MORNA BA amp Opinion Leaders 3 HAS A Relations is_linked_to Total nominees 25 Enabled nominees 25 Active nominees 24 Total nominations 120 AVG nominations nominee 4 750 STD nominations nominee 0 433 Repeat nominations Self nominations 9 Total leaders 5 Group max size 4 Ga a a n a i a a i a a a a a a att ak aa Leader In Degree Rel Integration after handling repeat and self nominatior 6 10 82 290 12 10 82 290 16 81 250 19 80 210 20 80 210 Group 1 leader 6 Group member Distance to the leader 1 1 Be 1 21 1 lt Select group lo El El Nodes 25 Links 120 5 nodes selected 12 links selected Figure 79 Opinion leaders selected nodes 4 The algorithm derives 5 leaders nodes 6 12 16 19 and 20 Their respective members are shown in the Opinion Leaders form under Group Leader and Group Members 5 As with other procedures the user can cycle through and select each of these derived leaders and members for closer examination using the Select group box on the form 6 The results can be saved pri
92. larity dissimilarity measures from step 2 and selecting a cluster or partitioning of the actors into positions 4 Superimposing the partition from step 3 on the basic input data and examining the positional or sub group density matrices to help interpret the relationship between and within the subgroups positions 5 Dichotomize the entries in the subgroup density matrices by using an alpha threshold to create a binary image matrix that indicates the presence or absence of positional relations To run the Roles and Positions procedure follow these steps 1 Open a graph a Note Any single graph with unweighted edges is supported in the current version of VisuaLyzer Future versions will support multiple and valued relations b VisuaLyzer will symmetrize the graph automatically using the maximum procedure if the graph is directed and will transform it into dissimilarities before submitting it to the roles and positions procedure for partitioning into discrete positions 2 Select Analysis gt Roles and Positions from the menu This will return the positional profile on the graph Fig 76 3 The Number of positions box on the Roles and Partitions form allows the user to select a number of positions of interest The example below selects three positions for the Krackhardt s advice network 4 Managers in each of these positions are structurally equivalent Membership of each of these positions can be interpreted based on knowledge
93. lation label at the top of this screen then type in the name of the new relation and click OK button This process can be repeated several times as required Figure 13 shows four 20 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Figure 13 Multiple links between nodes links between Node_001 and Node_002 Each of these links can have their own properties and attributes Once multiple link types have been created between two nodes each is selectable in the Relation drop down list box on the Edit Link screen when one of the links between those two nodes is selected Importing Data VisuaLyzer accepts data input from a variety of file types and formats to allow you to easily use the VisuaLyzer functions with your existing data Each of these data types is described below UCINET edgelistl edgearrayl files Both EdgeListl and EdgeArray1 files can be exported from UCINET The edgelistl format belongs to a family of linked lists format where a user specifies only ties that actually occur between dyads and omits those that did not occur from the data The edgelist1 format is typically used for square 1 mode adjacency matrices It usually takes the form ego alter relationship value That is it consists of a list of edges or links columns 1 and 2 followed by their values column 3 as shown in the example below dl n 3 format edgelistl labels Paul Liz Jane
94. lculating the value of Q as it goes along making it especially simple to find the optimal community structure we select the best partotion of the network by looking for the largest value of Q Finding clusters To implement clustering algorithm 1 Open a graph e g Krackhardt s Advice relation is used in this case 2 Select Analysis gt Partitions Clusters gt Clustering menu This will perform a hierarchical clustering procedure on the graph ese Visual yzer 1 1 Default Style Internal KrackHighTec eng DER Analysis Reasoning Help Network Properties yo P rey e amp a oy FI D 8 a E E Gh Node Centrality Adjacency Matrix Partitions Clusters Nearest Neighbors 3 A a Shortest Path P Sy i 16 E ada peck Cliques X artition levels Partitions Ch 3 Boo 3 ATN Partition level 3 Partitions Clusters gt Hierarchical Clustering N 3 FIA IModularity 0 0375 Roles and Positions Optimal Partition Communities f pl o arado dol Opinion Leaders VA BP J A Pe Cluster 1 7 E o 1 Core and Periphery g 1 Bact AR N 4 10 l A Z HUN A T Cluster 2 7 13 114 2 1S 6 8 9 Cluster 3 7 12 Number of clusters 3 Select cluster lo Show cluster ID Nodes 21 Links 190 Use Dbl click to add or edit a nodeflink right button add linkfaccess command menu left button drag select disable node Figure 74 Clusters The main output of the procedur
95. lected Moving a node To move a node click on the node and drag it to a new position on the screen Nodes can also be frozen or aligned These issues will be discussed later The current X and Y coordinates of a selected node will be displayed in the VisuaLyzer status bar at the bottom of the screen Display Properties VisuaLyzer provides several options for changing the way your graph looks To view some of these open the Display Properties window Fig 8 you can e press F4 e or select View Edit from the Display listing in the Edit menu e or click the display properties button Fig 7 14 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Display Properties General Nodes Links Nodes Selected 2 Color Shape Image 3 a Change 14 to 28 Eltrensp Size pixe 14 E Label Orientation Right Font AaBb Links Selected 1 b Click the General tab Color En wWidthnoixels 1 2 Style Font AaBb Arrow Head Color a Head Length pixels 110 Head Width pixels 3 Save Close Figure 8 Changing display properties This window was displayed with the Nodes Links tab showing because Node_015 and Node_001 were highlighted when we opened it This tab lets you change the ways nodes and links are displayed For example we can change the size of the nodes by changing the number in the Size field Change this number from 14 to 28 The size o
96. linked pairs query will accomplish this and display the results which can also be exported as similarity matrix 77 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved So let s take the Krackhardt advice network Our interest is in finding out age difference or asymmetry in advice relations 1 e the extent to which people seek advice from others who are older or younger than them To execute this query load the Krackhardt s advice network and select Linked Pairs Comparison from the Query menu This will bring up a window like the one shown below Linked Pairs Comparison Select link if the Following conditions applied to the difference between nodes that connected by the link Example AGE gt 10 Select ink Y the difference between AGE attribute values of endpoint nodes exceeds 10 years Attribute Function Diff Value Diff Value2 Age gt 20 Similarity Matrix 3 Set the conditions Figure 65 Linked pairs comparison Select an attribute of interest from the drop down box under the attribute column here Age Select a mathematical function e g less than greater than equal to etc here greater than since we want to find the difference of 20 years Set the comparison or difference value for the attribute here 20 years You can perform Boolean queries based on multiple attributes In the example a
97. members of the group do not have the same value for an attribute the value is displayed as a symbol as above You can add new attributes by clicking the New button and then typing in the name and value if any into the grid cells You can delete an attribute by selecting its row and clicking the Delete button To add or remove links to the group select the Members tab and use the corresponding buttons Attribute based display options Visual yzer lets you automatically change various node and link visual properties based on the values of their attributes This is especially useful when you have many nodes and attributes to color and don t want to do each individually These features make it easy to change the graph shown in Figure 50 to the more meaningful graph shown in Figure 51 62 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Daniel James Sarah Abigail David A Elizabeth Samantha Nicholas Alyssa John annah mily Christopher Joseph Jacob tauren f j gt Alexis a Matthew Figure 50 Basic graph 63 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ee Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng DER Fie Edit Layout View Query Analysis Reasoning Help 1208 ARMS AQ SCH BOOMS PPr AA HWHNA4 HOH
98. ministic subnetwork outdegrees lt 1 R complement R di di complete R Introductions Endpoints gt a IB was_infected_by Start End Links gt m he Self loops gt o Symmetries gt My Pi l Asymmetries gt a ah Completions k A Unique Links gt f b pr Rename Apply Colors Tests gt A 4 Select All Unselect All n Arbitrary Relations A ge X Cancel Operation 0 Pi J Yv ia fe Select attribute value s 1 oo Select All Unselect Close Figure 83 Finding the unknowns NA within a deterministic subnetwork Nodes 16 Links 12 The subnetwork limited to this known community that is functional has at most one output For The deterministic and nondeterministic components of the network are dual More generally a relation R is the sum of deterministic and nondeterministic parts The following equation expresses this fact R R complement R di R R di This equation has expressions that can divide edges and source nodes into two subsets Source nodes that link to only one sink belong to the deterministic part The remaining edges and source nodes belong in the nondeterministic part When a source node is NA and we restrict NA to those nodes in the known community in R then NA will take on every node in R The further restriction of nondeterminism selects some of those nodes and the deterministic constraint selects the remainder T
99. mk directions ignored Adi a ab ab i a a Path 1 16 2 14 Path 2 16 18 14 Path 3 16 9 14 Y Select path Nodes 21 Links 190 3 nodes selected 2 links selected Figure 72 Shortest path References e M E J Newman Scientific collaboration networks II Shortest paths weighted networks and centrality Phys Rev E 64 016132 2001 http www santafe edu mark papers 016132 pdf 87 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Cliques and Clusters Network Sub Structures Sub Groups Sub structures are important in network analysis because they enable an analyst to gain a micro view of the network showing the various groupings that may be present in a network and the implications of the sub group configuration for behavior of members For example conflict might be less likely in overlapping cliques where some people are members of both cliques because of shared mutual interests than in non overlapping ones and one might expect diffusion of ideas practices and infections to accelerate between such sub groups with overlapping memberships Knowledge about the location of an individual in a sub group may also be critical to understanding her his behavior in the network For example while some individuals may act as bridges cut points boundary spanners between groups others may have their connections contained within their own gro
100. mples http www mycompanywebsite com docs myresume html C photos mycompany me jpg Clicking the Open button will open the page or document using the default viewer program A node attributes and values will be displayed as a yellow tool tip when the cursor is held over it These attributes can also be permanently displayed for selected nodes by clicking the Toggle Attribute Windows toolbar button or hitting F3 Fig 11 18 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved ee Visual yzer 2 0 Default Style Internal Status Edgelypes_M eng File Edit Layout View Query Analysis Reasoning Help SAB HRS 12099 ROOMS PLACA WH aly Fred Bob Status 0 02 Se upal tie Toggle node attributes E Sharon FAkathy in n or off Status 0 02 Status 0 08 windows on or off Title auditor Pe Title secretary Stuart 3 A EN Status 0 19 Title supervisor 7 Susan ti Charles LE Es Soe anuel Fi oof Status 0 57 Doma te Title manager w Harold Py Wynn g Carol Nodes 14 Links 24 6 nodes selected 6 links selected Figure 11 Toggle node attributes windows Link Types and Attributes Links can also have attributes similar to nodes To enter and edit link information right click on a link and select Edit Link from the pop up menu This will bring up the Edit Link screen Fig 12 Double clicking on a link will also brin
101. multiple links between two nodes each must be moved individually Spring Embedding Layout VisuaLyzer will also automatically organize your nodes using a spring embedding layout To use the spring embedding layout either e click the Spring Embedding Layout button e select Spring Embedding Layout from the Layout menu e press the F9 key Circular Layout VisuaLyzer allows you to automatically organize your nodes using a circular layout To use the circular layout either e click the Circular Layout button or e select Circular Layout from the Layout menu Radial Layout VisuaLyzer will allow you to view your nodes according to their proximity to the center of the network To organize your nodes using a radial layout either e click the Radial Layout button or e select Radial Layout from the Layout menu Layered Layout VisuaLyzer also allows you to organize your data according to various data values This means that nodes can be placed according to values along an X and Y axes for example nodes with higher values can be placed at the top of the screen while nodes with lower values can be placed at the bottom Using the provided example file Tree with EdgeAttributes Types M eng you can load the Layered Layout setup screen shown below by either e clicking the Layered Layout button or e selecting Layered Layout from the Layout menu Layered Layout Select node attributes to use in Layered Layout Y axis Arrange laye
102. n Background color You can change the background color for a graph by clicking on the square to the right of the Background Color label at the top of this screen This will open a Select Color screen Click on a colored square and the background of your graph will change to that color 49 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Background Image You can also use an image for the background of your graph To choose the image click the rectangle to the right of the Background Image label at the top of the General tab of the Display Properties screen This will bring up a standard Windows open file screen from which you can pick your background image file Currently VisuaLyzer allows you to use bitmaps and JPEGs as background images Note that the aspect ratio of the image is preserved i e it is not stretched to fill the screen and is centered in the display If you have also chosen a background color it will be displayed it whatever space not taken up by the image To remove the background image click the blue word Clear beneath the Background Image label You can also hide and unhide the background image by un checking or checking the Show Background Image checkbox in the middle of the Display Properties screen Node color You can control the color of nodes created when you double click on the screen by changing the color of the box to the right of the Node Color label at th
103. n ii A aida A O A 88 Clusters and ES e ATARI RAR O O N E qu san cad 89 GATS Clusters iio ias 90 Finding COMMUNITIES minita dia iia E R AENG S 91 ROLES AMG Posts a mean e a a a aa a a N A 93 CUTPOIINS ii A a A A a a 95 Opinion Led ds 97 Algorithm Implementation is a 98 E A unt r meer ane Ae CENT east AP en AOS OMA Eee OEM E AER NE eM SR APN oP oe ee ee 100 FORMAL REASONING ABOUT SOCIAL NETWORK DATA oconccccconccoononsononnaconnncnonancononaccccnncnnnanccnonacos 104 EimkKsias Hacia A o at AC nh o e e e lt e Ale Eee eto dee de 104 Boolean Operations iria 104 e A A EI 105 A nae cated ease Rte OL tec ita iu ute Oe Gide Rah Gated hte OC adil ces ES 108 Solving for Unknowns in Deterministic Networks ceesseceesceceeseeceseeeceeeeeceeneeceeneeceeeeceeeeecseeeeeeeeenaeeeees 109 DRGs VG o A A a a a a a A cot Nols tect te has hee a cues 112 Exploring large hierarchies A A 115 Limitations and further informatio neeo is 116 Further IntormatiK oi lada 117 Appendix Mouse and Keyboard Commands cssscssssccsssccsssccsessscsssccesseccssssccessaccsssssssssescssssccssssscosees 118 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved TABLE OF FIGURES A A a ne th ice lot ad ints alt ac eth Bot so oe paral aS 9 Figure 2 Creating a random graph ii A EEES tae ee 10 Fig re zA tandon A A I NE a hg aa ae NE aia 10 Fig re4 Sprinp embedded layout sce 11 Proure 5 Circular layout usina ocn pd
104. n one bit to words of bits This in turn corresponds to an arrangement of switches in a parallel circuit As an example consider the subset a c d of the set a b c d This could be represented as a 4 bit word as 1011 where the left to right position of the bits correspond in alphabetic order to the letters a b c d Similarly this can match with 4 independent switches that are in turn closed open closed and closed Relations are sets as well sets of ordered pairs Bits O and 1 can still indicate presence or absence of a link and sequenced into a word as before to represent a set of links a relation Now a closed switch in a digital circuit can link two junctures two nodes when the switch is open the link is absent The set of edges in a network is a relation but some relations do not have a network representation For example the subset of edges in a network consisting of self loops is a subset of the identity relation Everything else the complement of the identity is the apartness relation that even relates nodes that are not necessarily in the original network The identity and apartness relations are infinite but implemented finitely as constraints Finally the nodes of the network themselves can have a deeper term structure In the reasoning part of VisuaLyzer the universe is limited to terms A term is a constant or variable NA or a complex term A complex term is a function symbol f a constant followed by a parenth
105. ng the relative sum should use amp instead of lt ampersand gt Further Information For more information write to Paul Broome net or Paul Broome MDLogix com Relational reasoning and algebraic logic are described in more detail in Broome P and J Lipton Constructive Relational Programming A Declarative Approach to Program Correctness and High Level Optimization Ninth Army Conference on Applied Mathematics and Computing 1991 Broome P and J Lipton Combinatory Logic Programming Computing in Relation Calculi International Logic Programming Symposium M Bruynooghe Ed MIT Press 1994 Jonsson B The theory of binary relations Collog Math Soc Janos Bolyai 54 Algebraic Logic Budapest Hungary 1988 245 292 Tarski A and S Givant A formalization of set theory without variables AMS Colloquium Publications Vol 41 AMS Providence R I 1987 Wasserman S and K Faust Social Network Analysis Methods and Applications Cambridge University Press Cambridge UK 1994 Broome P and J Lipton Combinatory Logic Programming Computing in Relation Calculi International Logic Programming Symposium M Bruynooghe Ed MIT Press 1994 Broome P Relations and Constraints for Constructive Programming SSGRR 2000 International Conference on Advances in Infrastructure for Electronic Business Science and Education on the Internet Scuola Superiore G Reiss Romoli L Aquila Italy Aug 2 2000 117 VisuaLyzer U
106. node gt lt node id Steven gt lt node gt lt nodeset gt lt nodes gt lt networks gt lt graph id VisuaLyzer sourceType agent targetType agent isDirected true gt lt edge name is_linked_to source Manuan target Allen type double value 1 gt lt properties gt lt property name Caption type string value Link_001 gt lt property name IsDirected type binary value false gt lt properties gt lt edge gt lt edge name is_linked_to source Allen target Steven type double value 1 gt lt properties gt lt property name Caption type string value Link_002 gt lt properties gt lt edge gt lt graph gt lt networks gt lt MetaMatrix gt lt DynamicNetwork gt 26 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Importing a DyNetML file e A DyNetML file can be prepared with any text editor or specialized xml editors according to the format described above e The file must be saved with a regular xml extention xml e Ensure there are no characters or spaces before the first line Xml declaration line to ensure format recognition Example of the line is lt xml version 1 0 encoding UTF 8 gt e To import a DyNetML file select Import from the File menu This will display a standard Windows file dialog like the one in Figure 15 page 24 Select DyNetML files in the Files of type box to ha
107. nted or exported to Excel and different colors can be used to differentiate the groups Algorithm Implementation The opinion leader algorithm discussed above assumes the graph is directed The following notes explain the steps and further assumptions underlying the process of identifying opinion leaders 1 Suppose we have survey data such that each of N persons nodes nominate no more than M persons from the network including possibly himself on a specified relation as his her adviser mentor or person he listens to or gets opinion from It attempts to a Find out who are the Opinion Leaders and b Assign match the rest of the network to the leaders optimally creating groups of approximately same size 98 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved 2 The number of leaders is derived from the pattern of nominations out degrees in the network If the SD AVERAGE ratio for the number of nominations per nominee is relatively small less that 0 67 in our case the program assumes there are M round AVERAGE maximum number of leaders in the network and suggests this number to the user to choose the exact number lt M that is desired 3 Ifthe ratio is big more than 67 the program can theoretically assign 10 of the network membership as leaders the assumption is that 10 would be convenient for most settings and applications Passive nominees nodes with out degr
108. o display his imported attributes VisuaLyzer 2 0 Default Style Default OfficePeopleList csy File Edit Layout View Query Analysis Reasoning Help 1208 HRS 210009 ROOMS LALALA WORA OGD els Edit Node Name Daniel Bp Type actor New User defined Attributes Attribute Value 54 Gender male Marry married Kids 2 Alyssa Nodes 18 Links 1 Node Daniel actor x 349 Y 169 Figure 18 Importing from CSV files 29 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Opening a File You can also open a previously saved VisuaLyzer eng file by selecting Open from the File menu This will display a screen like the one shown below Binary w Q 2 PE Ie 19hi_M eng Tree_with EdgeAttributes Types_M eng E 4ristotles_categories_M eng My Recent E Caffeine molecule_M eng Documents CakeHouse eng Es Co citation data eng Es Deedlit eng ls EgoNet eng E Family planning study eng E KrackHighTec_M eng E LanguagesFamily eng Es OpinionLeaders eng E Planar net_BkImage eng e Planar _100 nodes eng E Planar _200 nodes eng Ll File name Status EdgeTypes_M eng My Network Files of type Visual yzer files eng Figure 19 Opening saved VisuaLyser files Select VizuaLyzer files in the Files of type box on this screen to have these files displayed When you have found the file you
109. odes are enabled Number of the nodes that are enabled Disabled nodes are ignore parameters 80 are 21 nodes in the d in calculation of VisuaLyzer User Manual v 2 0 Isolate Nodes Groups Relations Current Relations Total links Current links Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Number of isolates in the currently displayed graph if any Groups are collection of nodes sharing similar attributes For example a network can be broken into two groups of male and female network members The Groups parameter indicates the total number of grouped nodes in the current network A group can contain 1 or more other nodes or groups For details on how to construct groups see section Grouping Nodes In VisuaLyzer a network graph eng file can consist of single or multiple relations or link types Operations including calculation of parameters can be restricted to any or all of these relations at any given time The Relations parameter indicates the link types in the graph even if they are not displayed or have been disabled For example the Krackhardt High Tech dataset contains three relations ADVICE FRIENDSHIP and REPORTS_TO but only the ADVICE relation is currently in focus and displayed hence it is indicated in the Current relations value Similar to the Relations the Current Relations parameter indicates the link type or relation under current considerati
110. of edges in the completion Similarly the network complement is the set of new edges An edge is in the graph complement of R if the edge is not in R 114 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Finally it is possible to test a relation R for totality by forming O R 40 This returns the universal relation if the original R was universal and O otherwise Dually we test for existence of any edge in R by forming 1 R 1 This returns the universal relation if there exists a pair of nodes in R and 0 otherwise Exploring large hierarchies Sometimes a network particularly an organizational or ontological structure is so large that an entire display has little value For example the NCI ontology is a hierarchy of concepts with approximately 33 000 facts that when shown in its entirety would likely not illustrate anything useful This is similar to the organizational display of a large corporation The NCI ontology is preloaded into VisuaLyzer at start time from the userdefined pl file as an example The layered layout of VisuaLyzer is configured by default to incrementally display a large hierarchy from the top In order to explore this hierarchy this top node must have one attribute defined called hierarchy For the beginning node hierarchy has the value 1 to indicate the top layer Additionally the large hierarchy must have already been defined as a binary relation
111. of node attributes might be correlated Note that if you have created grouped nodes see section on group nodes you cannot lay them out based on node attributes The group will stay in the same position However when grouped nodes are laid out according to node degree in a layered layout the node degree of the group is the number of external links to the group The grouped node will be displayed based on this number N mode Layout Use N mode layout to organize your data by node types This means that nodes are placed for example according to node attribute values along the Y axes and according node type along the X axes Nodes with higher attribute values can be placed at the top of the screen while nodes with lower values can be placed at the bottom Nodes of the same type will be grouped together in the same column To use the N mode layout either e click the N mode button or e select N mode Network Layout from the Layout menu This will display the N mode layout set up screen Fig 34 N mode Network Layout Select node attributes to use in N mode network layout Y axis Arrange layer nodes evenly along Y axis according to node degree O based on optimization algorithm min edge crossings Sort order Select order of node types x axis r o CTE actor Figure 34 N mode Network Layout options 44 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved
112. ogic Inc All rights reserved Network Animator Working folder E gt Rate 2 E frisec Cyclic playback Frame 0 of 0 Figure 62 Network animator Select the directory where the files are located by either typing the path in the Working folder field or clicking the button to the right of this field and selecting the directory from the Windows directory tree that is displayed Indicate the number of files to be displayed per second in the Rate field Check the Cyclic playback checkbox if you want the series of files to be displayed over and over continuously Click the Play button to play the sequence and click the Stop button to stop the sequence from playing Clicking the lt lt button will display the first frame in the sequence and clicking the gt gt button will display the last frame in the sequence Clicking the lt button will display the previous frame in the sequence and clicking the gt button will display the net frame in the sequence 75 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Data or Network Querying Queries enable selection of nodes or links that fit specific criteria set by the researcher VisuaLyzer implements two types of queries nodes search and linked pairs comparison Node search Node search queries find nodes or links that satisfy criteria set a priori by the researcher Let s take the Krack
113. ol as above You can add new attributes by clicking the New button and then typing in the name and value if any into the grid cells You can delete an attribute by selecting its row and clicking the Delete button 58 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved To view the Members tab click the Members tab at the top of the Edit Group screen This will display a screen like the one shown below de Edit Group Members Group Name Group_001 Nodes 18 Group members 4 Name A Name Abs Alexis Jacob Alyssa Lauren Christopher Matthew Daniel David Elizabeth Emily Hannah Jacob James John oe Figure 47 Edit group screen members Again you can change the name of the group by changing the value in the Group Name field at the top of the screen The list on the left contains all nodes and groups in the current graph The list on the right contains all nodes and groups that are members of the current group To add a node or group to the current group select that node or group in the left hand list and click the Add button or drag and drop the node into the Group members list To remove a node or group from the current group select that node or group in the right hand list and click the Remove button Link options VisuaLyzer provides several options for displaying links These options can be accessed and set by e Pressing the Di
114. on e g ADVICE relations This is the total number of links or edges or connections in all relations in the current graph regardless of whether relations are enabled or not For example all three relations in the Krackhardt High Tech data set produce 312 links Total number of links in currently selected networks or number of visible links This includes links connecting both enabled and disabled nodes a link could be enabled disabled by disabling its corresponding nodes Current Enabled links The Enabled links value indicates the number of the links that are enabled in the current Diameter Geodesic Distance Density network Disabled links are ignored during calculations a link is disabled if any of the connected nodes are disabled The longest geodesic distance within the network unless infinite The diameter of a graph is important because it quantifies how far apart the farthest two nodes in the graph are and therefore how long it would take to reach that node e g the furthest maximum distance a message would have to travel from node i to node j A pair of nodes could be connected by several paths or links which may differ in length A shortest path between two nodes is referred to as a geodesic If there is more than one shortest path between a pair of nodes then there are two geodesics between them Thus the geodesic distance between two nodes is defined as the length of the geodesic between them F
115. or valued graphs the geodesic distance is the cheapest or shortest path between points computed by summing the values of each link in the path Distances are relevant in social network analysis because they indicate how far apart nodes are Proportion of links that are actually present in a graph It is the ratio of total number of links to the maximum possible links For an all to all connected network Density 1 0 For valued graphs density is the average proportion of links or mean degree of nodes in a graph 81 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Centralization Overall network degree centralization measures the extent to which a single actor has high centrality and the others low centrality The larger the centralization score the more likely it is that a single actor is quite central with the remaining actors considerably less central and probably at the periphery of the network Essentially it expresses how unequal variable or heterogeneous the actor centralities are in the network It is expressed as a percentage For a star type network centralization 100 and for an all to all connected network centralization 0 Node Properties e g Node Centrality VisuaLyzer calculates a number of nodal properties Version 1 outputs 4 degree centrality measures They are shown below for the Krackhardt advice network Node Centrality Sh oS By The graph is
116. other nodes whereas peripheral nodes have weak connections with each other and with other nodes Ideal core periphery structure can be discrete or continuous Discrete models produce two distinct classes of nodes those of the core nodes that have relations with each other alone and those on the periphery The continuous model assigns nodes to different degrees of coreness a sort of a continuous measure Visual yzer implements a discrete model discussed by Borgatti 1999 In this model the core periphery analysis algorithm seeks a partition that maximizes density in the core and minimizes the density in the periphery Ties between these two regions are ignored in the computation A simple example of a symmetric adjacency matrix for an ideal structure is shown below In this case the nodes 1 2 and 3 are core nodes Note that because of ignoring some ties many networks satisfying this model will have the same core and periphery 1 2 3 4 5 6 7 e VisuaLyzer implements the Kernighan Lin algorithm that is proven to be fast and able to search global optimal core periphery structure see Boyd 2004 Boyd s description of the algorithm is provided below under Kernighan Lin algorithm e The algorithm begins by permuting the rows and columns of the adjacency matrix and comparing the results with some corresponding ideal matrix on each step It computes a measure of fitness that indicates how well the observed clusters approxim
117. ou have groups inside of groups and want all your nodes in all your groups to be ungrouped select Ungroup All from the right click pop up menu To view information about a group double click on it or select the group and then choose Group from the Edit menu This will bring up an Edit Group screen like the one shown below ee Edit Group ci Members Member Attributes Group Label IGroup_001 Node Type actor User defined Attributes Attribute GENDER Kids e MARRY lll group members Show values and apply changes to O top level nodes only top level nodes and nodes of all nested subgroups La JN Cancel Figure 46 Edit group screen This screen has two tabs Members and Member Attributes The Member Attributes tab shown above lets you change the label of the group by changing the value in the Group Label field at the top of the screen The grid in the middle of this tab displays the attributes and values for the members of the group If the top level nodes only radio button is checked data displayed and edited here will not be applied to nodes in groups that are in the current group If the top level nodes and nodes of all nested subgroups radio button is checked then data displayed and edited here will be applied to all nodes in all group within the current group If all members of the group do not have the same value for an attribute the value is displayed as a symb
118. ow if a network is deterministic or not because determinism indicates predictability and reproducibility More generally the relation R is deterministic if R R di 0 Since for any relation R R R R di R complement R di then we can divide the relation R into two parts a non deterministic part R R di and a deterministic part R complement R di More interestingly 1f the network contains unknowns these constraints will partition the unknowns accordingly and solve for missing nodes An example is the tracking of a contagious incurable disease infection Such an infection network has a converse network that is deterministic No one is re infected because no one is ever cured Medical science is complex and rarely this idealistic there are often exceptions to rules However computational systems are finite so a hierarchical description needs only further elaboration to discover such determinism in the network Of course the question is often whether this refinement is specific for the network or more importantly applicable to a population The has_infected relation links persons with those that the person has infected In the associated network each node has maximum in degree of one Thus the was_infected_by relation converse has_infected has maximum out degree of one and should be deterministic if the disease is indeed incurable The way to specify this as simple set operations is with the expression was_infected_by was_infected_by
119. p down list box and click the Save button VisuaLyzer lets you save your graph as any of the following image types Monochrome bitmap bmp 16 color bitmap bmp 256 Color Bitmap bmp 24 bit bitmap bmp JPEG jpg jpeg Windows Metafile emf Exporting data Along with saving your data as a VisuaLyzer file you can export your data to several different file types each of which is described below UCINET Edgelistl To export your data in EdgeListl format select Export from the File menu to display the standard Windows Save screen Type the name of the file to be saved in the File name field select UCINet EdgeListl in the Save as type drop down list box and click the Save button UCINET Edgearrayl To export your data in EdgeArrayl format select Export from the File menu to display the standard Windows Save screen Type the name of the file to be saved in the File name field select UCINet EdgeArray1 in the Save as type drop down list box and click the Save button GraphML To export your data in GraphML format select Export from the File menu to display the standard Windows save screen Type the name of the file to be saved in the File name field select GraphML files in the Save as type drop down list box and click the Save button DyNetML To export your data in DyNetML format select Export from the File menu to display the standard Windows Save screen Type the name of the file to be saved in th
120. positions 3 2 Select position o El Z Show position ID Nodes 21 Links 190 Use Dbl click to add or edit a nodeflink right button add linkfaccess command menu left button drag select disable node Figure 76 Roles and positions Density Matrix To obtain the density matrix for the advice relation click on the appropriate density matrix icon on the Roles and Positions form This will produce the matrix 1 2 3 Position 1 1 0 000 0 684 0 526 Position 2 2 0 684 0 000 1 000 Position 3 3 0 526 1 000 0 000 94 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved a The entry in the density matrix is the proportion of links that are present between the cluster in the row position and the cluster in the column position For example the entry 0 684 indicates that almost 70 of links between positions and 2 are in existence showing dense interconnectivity between the positions b Image Matrix To obtain the image matrix for the advice relation click on the appropriate image matrix icon on the Roles and Positions form This will produce the matrix Position 1 1 Position 2 2 Position 3 3 O m O N Oh rr Ww c The cell entries indicate the presence 1 or absence 0 of relations between the various positions or blocks Indicating the presence or absence of relation is based on setting a threshold or cut off value for the density matrix Cell entries
121. r nodes to the extent that other actors depend on it to make connections with other people For each actor the betweenness centrality routine calculates the proportion of times that they are between other actors e g for sending of information to arrive at a raw score for actor betweenness centrality This measure can be normalized by expressing it as a percentage of the maximum possible betweenness that an actor could have had The user should note that calculation of this parameter is a time consuming process as it includes finding all possible paths between j and k nodes Examples for 7 node network 1 Star network the center node has betweenness 100 and the six peripheral nodes 0 2 Circle network all nodes have the same betweenness 20 3 Line network the node in the center of the line node 4 has maximum betweenness 60 the two end nodes have minimum betweenness 0 References Freeman L C 1979 Centrality in Social Networks Conceptual clarification Social Networks 1 215 239 Borgatti S P Everett M G and Freeman L C 2002 Ucinet 6 for Windows Harvard Analytic Technologies 85 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Nearest neighbor The neighborhood of a node i in a graph consists of all nodes adjacent to that node The nearest neighbor s of a node are those nodes that are closest to it physically within a specific distance To find
122. raphml graphdrawing org 24 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved The GraphML format invariably consists of language core to describe the structural properties of a graph and optional extension components An example of the main language core and the resulting graph is produced below Example 1 Simple graph no attributes at all just nodes and edges lt xml version 1 1 encoding UTF 8 gt lt This file was written by mdlogix Visualizer 1 1 application gt lt DOCTYPE graphml SYSTEM http www graphdrawing org dtds graphml dtd gt lt graphml gt lt graph id Untitled edgedefault directed gt lt node id Allen Tien gt lt node id Manyuan gt lt node id Steven gt lt edge source Allen Tien target Manyuan directed false gt lt edge source Allen Tien target Steven directed true gt lt graph gt lt graphml gt Importing a GraphML file e A GraphML file can be prepared with any text editor or specialized xml editors according to the format described above e The file must be saved with a graphml graphml extention e Ensure there are no characters or spaces before the first two lines or Xml declaration lines to ensure format recognition Examples of the respective lines are e g lt xml version 1 1 encoding UTF 8 gt and lt This file was written by mdlogix Visualizer 1 1 application gt
123. reen and then pressing the Enter key Grouping Links VisuaLyzer allows you to set attributes and attribute values for groups of links To do this select the links you want to group then right click on the graph and select Group links from the pop up menu This will bring up an Edit Group of Links screen like the one shown below 61 VisuaLyzer User Manual v 2 0 or VisuaLyzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng File Edit Layout View Query Analysis Reasoning Help 1288 ams Copyright 2005 2007 Medical Decision Logic Inc AIOP BOOMS PLS CREM WHA AA Abigail 2 David e A Nicholas Alyssa f John Hannah E Emi Christopher sfe Edit Group of Links x Members Member Attributes lis_linked_to OvYes Ono Label Relation Directed Weight 1 000 User defined Attributes All rights reserved Attribute Co_workers Drink_partners N Friends o Joseph Other Sex_partners Jacob tauren eee po I Alexis Matthew Nodes 18 Links 17 4 nodes selected 3 links selected Figure 49 Edit attributes for group of links This screen has two tabs Members and Member Attributes The Member Attributes tab lets you change the relation weight and set the label for selected links The grid in the middle of this tab displays the custom attributes and values for the members of the group If all
124. right 2005 2007 Medical Decision Logic Inc All rights reserved Joseph O Alyssa christopher fe dd y y Emily J e ames Sarah Y o eS aiats ki erie Samantha Pa David John O Elizabeth Figure 21 Addition of nodes to existing graph Notice the two NewGuy nodes in the center of the graph connected to Hannah When merging graphs VisuaLyzer will first try to match the nodes in the new file with the ones in the existing file If they are an exact match based on the node s name and attributes they will be considered as the same node and only the node s links will be updated As an example we will start with the example file Square grid 5x5 graphml Fig 22 which has 25 nodes 32 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Dice 01 Node_07 s ESA Node_13 pues 02 aii d ha ia 08 f a t9 PE y i Al f i fa Tie Node Node_25 Node_03 j a E lode_20 Fi ex Node_09 r a Fi iiia a A dts j ye 15 7 ay 04 ps 21 Node 10 y E Node_27 i 05 pue 16 f i pus ri ee 11 Sgier 2 H E dian 28 Snes 17 y j ode 23 4 Node_29 Figure 22 Square grid 5X5 graphml aS wv o a sa f WP Node 02 Node_08 SN 5 ae Ss Ps DMode_06 _ ONode_03 ae a pat ie Oe yNode_07 Node 0i aen Fad m g Onide o5 Figure
125. rs according to value of GENDER O according to node degree Sort order Descending Ascending X axis Arrange layer nodes evenly along X axis O according to value of O according to node degree y based on optimization algorithm min edge crossings Default OK Cancel Figure 29 Layered Layout setup screen 40 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Nodes can also be organized along the Y axis according to the values of their attributes To do this select the radio button according to values of in the Y axis box in the Layered Layout setup screen and then choose an attribute in the drop down list box and click OK The result is shown in Figure 30 ee Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng DEAR Fie Edit Layout View Query Analysis Reasoning Help 1228 BAS kA Ow BOOMS PLE CAMA MERA BAGH GENDER iii d i WiP cue eel a Nicholas John 0 David 0 James male ES f E i x S DN H if Es Gp j de pS i f H j 5 a y Samantha M Elizabeth 0 Alyssa W Lauren Nodes 18 Links 17 Use Dbl click to add or edit a nodejlink right button add link access menu left button drag select disable node s Figure 30 Layered Layout graph 41 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved S
126. rs see node cluster section can be changed by changing the color of the box next to the Cluster Frame Color label at the top of the Select Color screen This color is changed like the background color is changed click on the box to open the Select Color screen and then click on a colored square to select that color Note that changing this color does not affect the color of existing cluster frames outlines Position Frame Color The color of the frame outline of node position role clusters described below can be changed by changing the color of the box next to the Position Frame Color label at the top of the General tab of the Display Properties screen This color is changed like the background color is changed click on the box to open the Select Color screen and then click on a colored square to select that color Note that changing this color does not affect the color of existing position role frames outlines 50 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Show Shapes Checking and un checking the Show Shapes checkbox on the General tab of the Display Properties screen will show or hide each node s shape This affects both all the nodes currently in the graph and all new nodes Show Images Like the graph background nodes can also have an image as well as a color and shape These options are set on the Nodes Links tab described below You can hide and unhide th
127. s then the answer returned from the solution procedure may not be displayed as expected This limitation will be removed in the future Selectors can also produce structures that do not match the arrangement of attribute attribute values at all The Reasoning part of VisuaLyzer assumes that nodes with identical names attributes and attribute values are identical This differs from other analysis parts of VisuaLyzer which consider distinctly drawn nodes to be different no matter whether they have identical names and attributes or not A name NA can refer to many nodes Note that a relation complement by itself can be expensive to compute but is not very useful in isolation Since it is almost always used in combination with other constraints this is rarely a problem In particular if a relation R has about 20 edges then complement R on its own is barely manageable on a desktop system while the intersection with a set S that is S complement R of perhaps 1 million edges can be straightforward An attribute has a value or the value is NA The set expressions in most general form can include inequalities that can state whether an attribute variable cannot have a value These inequalities are not represented as networks but instead are depicted with more general set expressions Avoid equational reasoning on floating point attribute values This is in part because floating point has always been approximate arithmetic that may give unintuitive r
128. s E E oe ee Aa E E ces aie E NE E ea EE E ERR ee aes 54 Elma ge iii 56 VA E EE E E E E O AEE 56 Label Ormentati onean ida 56 RE 56 Disabling INOGES uta sii 57 Grouping NODES siii dd iban 58 A NT E 59 CAI sa 60 A seasicdaesastyaietasnasienessdovacvashesesee vadydcadpacuatusaevadensopadei sodanvaaswavasta ogee E 60 A 61 A E a a O OS 61 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Arrowhead OL OE A A DS 61 E 61 O A A os a is a ie seek A A Gis co atts Bag 61 Grouping Ta AK ii id 61 AUDE DAS O od 62 A On a on ee ee eR ue eR Ce SOE aa Mee Ne ee PRT 64 Node Color iii A tee Ud ca eke ates neds aed eee 65 Node apta ta e Ds e e el 67 NOE A ese te cata od E N oh age A ca theatand anaes 68 TST BOLT Seas O E cae 69 Legend for attributesbascd que nd 70 VES A ii 71 Display Relatos PES e ras 12 Graph Animations O O 74 Data Or Network Quer ying anidar RA vst bo rd Esku S Esto N oS son E E Eoo SEESE E CSSS 76 INO GE SS ANCL se yeasts none i a e e a a aep i es poisas 76 Linked Pairs comparison nsee E A TE EEE E E A TEES 77 GHUGTY FAVS a cad 79 ANALYZING SOCIAL NETWORK DATA osiniid nico iii cd EA nda siii 80 Network Level Properties visit ae rn 80 Node Properties e E Node as 82 Nearest a A A A A UR Saal lan 86 Shortest Palin ia ad adidas 87 Cliques and Clusters Network Sub Structures Sub Groups ccsscceesseceeseeceeseeceeneeceseeecseneecseneecseeecseeeesaes 88 Clique Identificatio
129. s and interactions with other actors Similar relations doesn t necessarily mean directly connected though For example students in different schools have similar relations with teachers and parents They occupy the position of student and perform a specific role though individual students may not know each other Thus position and role is a measure of structural equivalence rather than adjacency or connectedness Structurally equivalent actors have the same profile except for the diagonal entries of the adjacency matrix Role and position algorithms often take as their input a similarity or distance matrix Some common measures of similarity and dissimilarity are Euclidean Distance and Pearson Correlation VisuaLyzer implements a single link hierarchical clustering using Euclidean distance as a measure of dissimilarity Starting with a state in which each node is the sole member of one of N clusters it repeatedly joins clusters together in pairs choosing at each step the actor that is closer to others by his her position in the network The running time of the algorithm is O N 2 Performing role and positional analysis involves the following steps 1 Preparing the data prior to analysis e g calculation of similarity and dissimilarities scaling if any 2 Selecting a similarity or dissimilarity measure which gauges the extent to which two actors approach structural equivalence or substitutability 3 Clustering the simi
130. s between UCINET and VisuaLyzer s current data specifications a UCINET identifies each relationship in a multi relational edgelistl dataset with a label and embeds the labels within the data what it calls level labels This is useful to distinguish batches of relational data However importing such a file gives a known error Number of data sections non consistent with level labels To get around this the user has to remove embedded level labels in the file this involves removing the main heading LEVEL LABELS EMBEDDED as well as the individual labels for each matrix The multi relational import proceeds well after this though users are advised to keep track of the names and order of the matrices in the edgelist file Future versions of VisuaLyzer will support the embedded labels feature 23 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Importing into VisuaLyzer To import files of this type select Import from the File menu This will display a standard Windows file dialog like the one shown below Look in UCINET QS em i i E Tree_with EdgeAttributes txt E 3D_Cube_symmetre txt E Tree_with EdgeAttributes Add txt My Recent E 3D_Octagon txt E Tree_with Edge Attributes AddNodeAttr Documents E 3D_Parallelepiped txt 3D_Pentagon txt E attributesTree_with Edgeattributes txt E Planar_fig31 txt E Planar _fig33 txt E Plamar_fig34 txt E Ring
131. s iaria ee Be ee eed 65 Figure 53 Attribute based selection screen color tab ssessseeseeseseseseessessresressessreseeseeseestssresesetestenseseresressetse 66 Figure 54 Selecting group of values Eolortabani o e e e dl eto 67 Figure 55 Attribute based selection screen shape tab 2 ih fieiek arene alan see anda 68 Figure 56 Attribute selection screen SILO ee Ai en GR Si ee ee a 69 Figure 57 Attribute based selection screen link color tab soriana 70 Figure 58 Legend window vii o ee cee 71 Fig re D9 Oacime and Savins Vie WS aa caad cosemeanet vaaeaseauaaas sup aTa meneeaea eceaueareneess 12 Figure 60 Select relation ksa A E uate ASE 73 Figure 61 Select typeof mode da 74 Fig re 62 Network animator ssc cost dest tose cia thee sto 75 Fig re 63 Nod searehs cainnean ala ec ee an ee E E E EE A a heeded 76 Fig re OF Re s ltof o O 11 Fig re 65 Linked pairs COPA sosea sereia iosia i a aeieea 78 Figure 66 Results of linked pairs COMPAaLISON secceeeseceesseceessecesaecessaeceeaeeceeaeeceeaeeceeaeeceeaeeceeeeeceeeeecseeeeceeeeees 79 Fig re 67 Network Properties 0 A A A da 80 Figure 68 Node properties degree centrality ai A A AA 82 Figure 69 Node properties ClOSENESS isa iii diria 84 Figure 70 Node properties DE Were e ae le LS 85 Figure 71 Nearestnereh bors a A A A Het tas anise 86 Preune 7 2 SHOMESE path A ee lad al a ah E SS cl 87 EA A a a i Ged tas a asi 89 Fig re 74 Cluster sio pd dada E Ei aS 90 Hegre 73 COMMUNIE eai o
132. se Nodes 14 Links 24 Use Dbl click to add or edit a nodeflink right button add linkfaccess menu left button drag select disable node s Figure 48 Display properties screen links This window can remain open and on top as you work with the VisuaLyzer display For example you can create and drag links while this window is open When you select a link or links when this tab is open the information on the tab will be updated to reflect the display information for the selected link s Color You can change the color of the selected link s by clicking on the square to the right of the Color label in the Links section at the bottom of the Nodes Links tab of the Display Properties screen This will open a Select Color screen Click on a colored square and the selected link s will change to that color Width You can change the width or thickness of the selected link s by entering a number representing a number of pixels in the field to the right of the Width pixels label in the Links section at the bottom of the Nodes Links tab of the Display Properties screen and then pressing the Enter key 60 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Style You can change the style of the selected link s by selecting a line type in the drop down list box to the right of the Style label in the Links section at the bottom of the Nodes Links tab of the Display Properties
133. seek advice of two intermediates who in 112 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved turn seek advice from one individual at the top This top individual is the one discovered as a position leader and can be compared to the group of three that are determined to be the default opinion leaders If a single opinion leader is specified only node 1 is found If two opinion leaders are specified then nodes 1 and 2 are found Finally four opinion leaders are nodes 1 2 3 and 4 These opinion leaders are not pure leaders as they have no followers ete Visual yzer 1 1 4PuzzlingOpinionLeaders eng Opinion Leaders File Edit Layout Query Analysis Reasoning Help 3 N o A DR En A a n y gt bl gt ES Se nl fe e Gt amp 4 Relations seeks_advice_of PRIOR position A Total nominees 7 a Enabled nominees 7 Active nominees 6 Total nominations 20 AVG nominations nominee 3 000 i 5 STD nominations nominee 1 414 SEE 4 mes Repeat nominations 2 Select Relation Links f gt Self nominations 0 f GS Total leaders 3 E seeks_advice_of F Ss Group max size 2 M PRIOR position leaders seeks_advice dh EE AS Y Leader In Degree Rel Integration after handling repeat and self nominations 1 3 50 000 2 3 50 000 3 3 50 000 Group 1 leader 1 Group member Distance to the leader Apply Colors A 4 1 6 Select All Unselec
134. ser Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Appendix Mouse and Keyboard Commands Mouse Right Button Click popup command menu Mouse Right Button Click Drag Release add link Click on from_node Release on to_node or on empty space on canvas if to_node not exists it will be created and connected to from_node Click If click on empty space on canvas unselect all else select node or link link is selected if nodes on both ends are selected Click on node link label does the same Click Drug If click on node move selected node s else if click on link move selected link s together with connected nodes else if click on empty space select number of nodes links group selecting Drop node s on group node Add selected node s to the group Double Click If click on empty space on canvas add new node else edit selected node or link properties Shift Click Select another node or link while preserving other nodes links selection Ctrl Click Disable Enable selected node Alt Click Freeze Unfreeze selected node position Fl Context help currently under development F2 Edit node link properties whatever is selected F3 Toggle node attributes windows On Off for selected nodes only F4 Edit display properties appearance and behavior settings depends on node link selection F5 Add new node F6 Straighten out links F7 Add new link endpoint nodes have to be selected F8 Create gro
135. splay Properties button e Pressing the F4 key e Selecting Edit gt Display Properties gt View Edit from the main menu This will bring up the Display Properties screen To access the Nodes Links tab click on the Nodes Links tab at the top of the screen see Figure 41 Also note that this screen reflects the information for the currently selected link s No links were selected when the above screen shot was taken thus there are no values displayed No value information is also displayed when multiple links are selected and have conflicting values 59 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved e g they are different colors or widths When a link is selected the Nodes Links tab would look something like the screen shown below eze Visual yzer 2 0 Default Style Internal Status EdgeTypes_M eng File Edit Layout View Query Analysis Reasoning Help ALSP BOOM PAOLA 1908 0433 o General Nodes Links Nodes Selected 0 C Transp Clear Size pixels Label Color Fred Shape BB Image x N B Sharon M Kathy i A PS PO Orientation 4 Font JO Susan A Links Color Width pixels P A A oe pe 9 Manuel a E e Charles _ Donna Style Carol pe Font PH Arrow Head Color a ey A Head Length pixels 10 l Harold Head Width pixels 3 W Load Settings Save Settings Default Clo
136. t All Link attribute Group 2 leader 2 Group member Distance to the leader 7 Group 3 leader 3 Select attribute value s Group member Distance to the leader 5 Select All Unselect All j l Select group lo y f Nodes 7 Links 20 3 nodes selected 6 links selected Figure 85 Links to Position leaders are in black Opinion leaders are in red It is sometimes important to impose self reference and sometimes to disallow it When sociomatrices are the underlying representation this corresponds to the diagonal However a logical representation includes self reference as simply the identity relation id x yl x y and excludes it with the apartness relation di the complement id So given a relation R the diagonal is included with R id and excluded with R di In the theory of relations and as should be apparent from the two previous examples there are two kinds of duality absolute Boolean and relative Peircean Any relational expression has a truth symmetric form derived by transformation using the relation complement In addition there are equations in the relation algebra that connect these forms For example the equation complement R id complement R di roughly says that we can exclude the diagonal from the complement R by including the diagonal in R and then complementing the result The deterministic nondeterministic networks also illustrate the
137. t eng File Edit Layout Query Analysis Reasoning Help SaaS G Semin Pe See MO HPN A Nondeterminisms gt 0 1 Introductions 10 dE 3 Endpoints t 13 EFI __ E Milis_friends_with Start End Links gt 14 O M seeks_advice_of gr oai E Wi reports_to Self loops gt A b Ne M reports_to complement seeks_ad Symmetries 1 15 9 Asymmetries gt S i 3 be 17 P 5 Completions gt 4 m z Pd Unique Links r 6 m 16 Rename Apply Colors Tests 10 8 _Select Al nh Ms Bi Arbitrary Relations Relation Combinators gt X lt Cancel Operation Execute Relational Expression Ctrl R Nodes 21 Links 21 Enter and execute an expression representing a new relation Figure 81 Determine who reports to but does not seek the advice of which manager There is only one edge only two nodes in the reports_to complement seeks_advice_of relation In this case it is true that reports to 78 but not true that 77 seeks the advice of 1 8 Note that if animation is turned off then the new link the new relation reports_to complement seeks_advice_of is drawn without disturbing the original relation reports_to This is done with F4 General properties gt General gt uncheck Animation Variables The relative product fann bill 1 has a set description which requires the use of variables As a set it is x yl x ann the variable y can have any value as it is uncons
138. tation You can change the label orientation where the label is in relation to the node of the selected nodes s by selecting a choice from the drop down list box to the right of the Label Orientation label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen Font To change the font of the label of the selected nodes s click the letters or blank space to the right of the Font label in the Nodes section at the top of the Nodes Links tab of the Display Properties screen This will display a screen like the one shown below _ Font style Bold Italic 1 1 m 8 Small Fonts A Regular Em Tp Souvenir Lt BT Italic 120 Tr SPSS Marker Set Bold Staccato222 BT EXE T Swiss911 XCm BT F Sydnie O Sylfaen Cancel Effects C Strikeout C Underline Color l GE Script Westem Figure 44 Font selection screen Select the options you want for the font and then click the OK button to change the node label s 56 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Disabling Nodes If you are interested in seeing what your data would be like such as layout or network analyses without a certain node or nodes you can do so without having to delete the node s and then re enter them again later When a node and its links is disabled it is not used for determining layouts or other analyses discussed below
139. ted the Nodes Links tab would look something like Figure 41 53 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved se Visual yzer 2 0 Default Style Internal Tree_with EdgeAttributes Types_M eng File Edit Layout View Query Analysis Reasoning Help ZB ams AJP BOOM SPSL ERM WERA Baw n oM General Nodes Links Nodes Selected 1 Color E Shape O Image C Transp Clear Size pixels 16 Label Fay E Orientation Right Sa Font AaBb Links Selected 0 Color width pixels fa Style Font Daniel dh James Y j ss Abigail A David a j X O Eizabeth Samantha Nicholas A A Y Alyssa A a John Ng A Wena Gin 0 Christopher f X j 7 5 Joseph ji Arrow Head Color Jacob P y tauren f SN o Alexis e Matthew Head Length pixels Es Head Width pixels fal ly Load Settings Save Settings Nodes 18 Links 17 Node Sarah actor X 545 Y 147 Figure 41 Node properties on display screen Again note that this window can remain open and on top as you work with the VisuaLyzer display For example you can create and drag nodes while this window is open When you select a node or nodes when this tab is open the information on the tab will be updated to reflect the display information for the selected node s Color You can chang
140. that exceed the cut off value indicate the presence of a relation between the positions and zero otherwise Traditionally researchers have used the overall density alpha density of the relation matrix as a cut off value VisuaLyzer implements a rather simple cut off value of 0 indicating any density value above 0 as indicating the presence of relations This may over estimate the presence of relations between positions and future versions of VisuaLyzer will allow the user to select a different threshold value to create the image matrix References e S Wasserman and K Faust Social network analysis Methods and Applications Cambridge University Press Cambridge 1994 e Burt R 1976 Positions in Networks Social Forces 55 93 122 Cutpoints In social network analysis cutpoint or 1 node cutset is a node whose deletion would increase the number of components in a graph The graph may contain many just one cutpoint or even no cutpoints at all consider a circle two nodes have to be removed to disconnect the graph VisuaLyzer recognizes isolated nodes dyads and components of size 3 and more These components are one of the results of partition cluster analysis So to find cutpoints VisuaLyzer runs partition cluster analysis for virtual removal of each graph node and compares the total number of components before and after the removal As always disabled nodes as well as isolates and nodes with degree centrality of 1 are exclude
141. trained In VisuaLyzer this would be depicted as a single link from a node named ann to a node named NA Not Available In general a node named NA can have any number of valid values for each single depiction As another example consider fann bill di which is the set x ylx ann y bill In this example y can again take on a number of values everyone but bill This set is not depicted in VisuaLyzer as a network as attributes in VisuaLyzer can only be positive equality constraints Instead this is displayed in another window as a set description Other representation examples based on the knowledge encoded in the relations is_friends_with seeks_advice_of and reports_to follow The italicized words e g Friend Advisor Someone Who and Whom replace the object variables x and y in the corresponding set expression 1 Who seeks advice from which Friends seeks_advice_of is_friends_with 2 Who seeks advice from which Friends that are also supervisors seeks_advice_of is_friends_with reports_to 3 Who has Friends_of_friends that are not friends i e Who are Candidates for introduction is_friends_with is_friends_with complement is_friends_with di 108 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved 4 Who has an Advisor that is also either a friend or a boss or both seeks_advice_of is_friends_with reports_to 5 Who has no subordinates reports_to 1
142. unication patterns of leaders and their followers VisuaLyzer implements one technique a matching algorithm developed by T Valente 1996 1990 to select opinion leaders and match them to their followers The algorithm assumes directed relations It attempts to e Find out who are the Opinion Leaders and e Assign match the rest of the network to the leaders optimally creating groups of approximately the same size A detailed discussion of the algorithm and its assumptions are presented in the Algorithm Implementation section below To run the Opinion Leaders procedure follow these steps 1 Open a directed graph The graph used in this example can be found at Examples Binary OpinionLeaders eng in the installation folder 2 Select Analysis gt Opinion Leaders from the menu This will first present the user with a pop up message box requesting the desired number of opinion leaders to derive up to the maximum detected and suggested by the program Leaders Number of opinion leaders s f The number of opinion leaders is derived from the pattern of nominations outdegrees in the network Depending on the size N of the network this number can vary from 1 to N though Fewer opinion leaders e g less than 5 make practical sense Further details about computation and assumptions are available in the user guide Figure 78 Selecting number of leaders 3 Select the number of opinion leaders desired and click OK button Th
143. up F9 Spring Embedding Layout start stop get organized F10 Bestfit F12 Toggle mouse mode magnifier regular Del Disable Enable selected nodes disabled nodes are ignored during calculations Ctrl lt gt Zoom In Ctrl lt gt Zoom Out Ctrl A Select all nodes if graph component has focus Ctrl C Copy selected nodes edges to Clipboard Ctrl F Find node by label pattern search Ctrl L Open Legend screen Ctrl N New file Ctrl1 O Open file VisuaLyzer binary eng file 118 VisuaLyzer User Manual v 2 0 Ctrl R Ctr1 Q Ctrl S Ctrl T Ctrl V Ctrl W Ctrl X Ctrl Z Run Relational Expression Builder Run search query Save file VisuaLyzer binary eng file Open Select Relation Type screen Paste nodes edges from Clipboard Load and apply a View Cut selected nodes edges Undo last operation 119 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved
144. ups while others may be isolated Such locational configurations may help the analyst examine extent to which their embeddedness may facilitate or impede transmission of information and diseases and diffusion of practices between the subgroups Network analysts have developed a number of measures such as cliques n cliques n clans and k plexes clusters and partitions that identify the existence of such groups and sub structures VisuaLyzer implements two of these measures cliques and clusters partitions Clique Identification VisuaLyzer implements the Bron and Kerbosch 1973 algorithm for clique identification The algorithm produces basic cliques with overlapping memberships meaning nodes can belong to more than one clique To generate a clique follow these steps 1 Open a graph 2 Select Cliques from the Analysis menu 3 This would open up a clique configuration panel as well as the graph showing the default cliques containing tree or more members by default VisuaLyzer produces cliques with at least 3 members 4 You can change the minimum size of the clique using the configuration panel and you can select a particular clique clique 1 2 3 etc at a time for closer inspection 5 Once you select a particular clique membership you can apply a number of attributes shape color and size to visually distinguish them 6 For example Figure 73 shows cliques minimum size 5 for Krackhardt s High Tech managers advice n
145. ute Based Query button and select the Color tab This will display the Color tab of the Attribute Based screen Select the attribute you want used to determine color in the drop down list box at the top of this screen The Legend box will then be filled with a list of the values for that attribute with a colored square next to each value indicating the color that will be used for nodes with that value Fig 53 65 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Attribute based Label Color Shape Size Link Color Node attribute MARRY Legend Wi divorced M married E single C Gray scale Figure 53 Attribute based selection screen color tab You can change the color used for an attribute value by e selecting an attribute value label or colored square to the left clicking the Select button e double clicking on label or colored square e using Ctrl Shift to select group of values e using collapsed values see figure below and select the new color from the color palette When the Collapse Values button is selected the resulting dialogue window allows you to specify the start value end value and the desired color for the group of nodes Fig 54 Note The preceding process can be applied to node shape node size and link color data as well To apply colors to the nodes on the screen click Apply Note that the Show Image box on the Gener
146. ve these files displayed When you have found the file you want to import either double click or single click on it for it to be displayed in the File name box and then click the Open button Your data will be displayed as a network graph Microsoft Excel files VisuaLyzer allows you to import data from Microsoft Excel workbooks even if you do not have the Microsoft Office Suite installed on your system To import data from an Excel workbook select Import from the File menu This will display a standard Windows file dialog like the one in Figure 14 Select Microsoft Excel Workbook in the Files of type box on this screen to have these files displayed When you have found the file you want to import either double click or single click on it for 1t to be displayed in the File name box and then click the Open button This will then display the screen shown below Select Excel Sheet Edgelist Adjacency Matrix Workbook C WisualizerlExamplesiExcelStatus xls Look linked node pairs in Sheet1 Edgelist Header row Look node attributes in Sheet2 Attributes Header row Figure 16 Importing from Excel Select the sheet with the linked node pairs in the top drop down list box and the sheet with the node attributes in the bottom drop down list box Check the Header row checkbox below the drop down list box labels if the 27 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved
147. xtent seniority level The algorithm produces two communities the high modularity index indicates a good fit to the data The algorithm as well as knowledge of the organization may provide further strength and validation of this observation This implementation would be most useful when applied to large networks which may a priori have some internal divisions References Borgatti S P Everett M G and Freeman L C 2002 Ucinet 6 for Windows Harvard Analytic Technologies Johnson S C 1967 Hierarchical clustering schemes Psychometrika 32 241 253 M E J Newman Detecting community structure in networks In Eur Phys J B 2004 http www personal umich edu mejn papers epjb pdf M E J Newman and M Girvan Finding and evaluating community structure in networks Physical Review E 69 026113 2004 http arxiv org PS_cache cond mat pdf 0308 0308217 pdf M E J Newman Fast Algorithm for Detecting Community Structure in Networks Physical Review E 69 066133 2004 http aps arxiv org PS_cache cond mat pdf 0309 0309508 pdf 92 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Roles and Positions In social network analysis the concept of position refers to a collection of individuals who are similarly embedded in network of relations while role refers to the pattern of relations between positions Actors in a same position have similar patterns of relation
148. zer provides several features and functions for customizing the display of your graph Moving nodes To move a node click on it and hold the mouse button down then drag the node to where you want it and release the mouse button All the node s links will follow the node to the new location After moving a node the links may not be lined up correctly they may be bent at odd angles as they follow the node you move While you can manually adjust the links as described below you can also have the program automatically fix the links e by pressing the F6 key e or by right clicking on the graph and selecting Straighten out links from the pop up menu Selected nodes can be frozen by right clicking on the graph screen and choosing Node s position gt Freeze nodes The screen position of the selected nodes will remain unchanged even when choosing a different layout Nodes cannot be moved until they are unfrozen Moving links Visual yzer also lets you control the curve of your links to help minimize line crossings To move a link first click it This will change its color and display a small drag square or control point Fig 28 aii Figure 28 Moving links 39 VisuaLyzer User Manual v 2 0 Copyright 2005 2007 Medical Decision Logic Inc All rights reserved Click on this square hold the mouse button down and drag it until the link is in the position you want Then release the mouse button If there are

VisuaLyzer User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents