ETACE Virtual Appliance User Manual
Contents
1. A Clicking the right mouse button orders the list alphabetically e Enter one value or multi ple by comma separated values in the text field Add Confirm by clicking Add or return e Select one or more values and click remove to remove these values from the list e Click OK to confirm all changes and close the dia OK log window Remove e Discard the changes and close the dialog window Discard These radio buttons can be used to define whether or not the databases should be compressed after finishing all jobs Depending on the storage settings compress ing databases can save hard disk space These radio buttons can be used in combination with the Do not run radio button to compress and decompress data bases without running the simulations again 4 How to use the Simulation GUI for customized Simulation Experiments 10 This table specifies which agent type should be recorded and at which frequency e By checking the check boxes in the Record col umn the user selects the agent type to be recorded Agent Record Period Phase s ai p e The user can determine the periodicity of data recording 20 by default which means the data is recorded on a monthly base and the phase shift 0 by default which means in combination with period 20 that data is recorded at iteration 20 40 60 etc e The snapshot option allows to record a full snap shot of the ag2. Eurostat Bank Government CentralBank ClearingHouse Time Series for Agent Firm Variable Name Method Filter 1 Filter 2 Further settin output output Mean vy No Filter vy No Filter 1 technology technology Mean vy No Filter vy No Filter Time Series to plot time series of sin mean_specific mean_specific Mean vy No Filter vy No Filter wage offer wage offer Mean vy No Filter vy No Filter gle variables price price Mean vy No Filter vy No Filter labour_costs_ labour_costs_ Mean vy No Filter vy No Filter 2 Multiple Time Series to plot multi ple time series in a common plot using the same scale 3 Agent distributions to create distri bution plots such as box plots or his tograms for specified iterations Transition Phase lo Add Legends Single Run Analysis v Batch Run Analysis Vv Colored Plots m File Type of Plots pdf x Parameter Analysis r 4 How to use the Simulation GUI for customized Simulation Experiments 13 The plotting process is automatically started after the simulations have finished However if the simulation data already exists the plotting can also be started without running simulations Therefore one has to select the Do not run Simulation radio button on the Simulation Settings tab In this case the sim ulation step is skipped and the plotting starts imme diately Run Simulations Do n
3. STATUS _itersdb STATUS_JOBS E top_level_folder v its ratio_data txt Configure r histogram_data txt heat_maps_data_2V txt gendb py launch_job_list sh E run sh GH compress_db sh B set_exp sh exp_script_1_setup sh B r_serial sh B create _job_list_compress sh B job list_O sh aT probability tech_policy_effectiveness Av 0 0 b run L gt run_2 L b run L bii run_4 L b E runs Bs b gt 14 WW top_level_folder gt pits b log LD E Plots E job_list_2 sh GH run_exp sh E job_list_3 sh B job_list_1 sh B exp_script_2_specific sh GH create _job_list sh experiment xml output xml environment xml shadow_model xml GlobalSettings xml e Before starting a simulation the simulation has to be built by clicking Build Experiment This means the configuration of the experiment which has been entered through the GUI is translated into files sh xml and txt files which can be read by the simulator to run the sim ulations the R software to analyze the data and by the operating sys tem for file operations and to control the sequence of activities e The GUI writes these scripts in the top level folder of the work space e Clicking Run Experiment launches the actual simulation experiment A simulation experiment is a two step process i The simulation of the model The simulator main ex ecutable is executed and writes the s
4. Set Paths item of the Settings menu to check or change paths used for the simulation The paths to be set are Set Model Executable Set initial Data File 0 xml e path to the main executable Set path to R Scripts Set path to xparser e path to the 0 xml file 4 How to use the Simulation GUI for customized Simulation Experiments 6 e path to the top level folder of the R scripts In the VA the full path of this folder is home eurace Desktop src JavaGUI src Data_Analysis_GUI_Serial e path to the executable of the xparser In the VA the full path is home eurace Desktop src xparser 0 17 1 xparser A For the pre configured experiments in Preconfigured Experiments all paths have already been set correctly 4 2 How to setup an Experiment The Simulation Settings tab which is the first tab on the main dialog window can be used to specify the settings that are related to the simulations There are basically three different kinds of settings to be can specified Simulation Settings Plotting Settings Simulation Settings Plotting Settings Run Simulations Don a ee 1 General setup of the simulations i e tuber of iterations 25000 the length of the simulation number of Nees anil aa iterations the number of batch runs and the parametrization of the simu lated model 2 Experiment mode i e whether or not the model is run multiple times with a specifi
5. lt gt ETACE Virtual Appliance wf User Manual Gregor Bohl Philipp Harting Sander van der Hoog Bielefeld University Chair for Economic Theory and Computational Economics ETACE April 16 2015 Contents Purpose and Overview 2 1 a Programmes Documentation and the EuraceQUnibi Model 2 Oinck Siarter Gide a ses raisa eee chad ea ec teen e edt ein eo KG How to use the Simulation GUI for customized Simulation Experiments Flame Modeling Environment 00 00 2 eee ee ee Lien serre Enron Leener eee eee ee ee eee eee HD oO KR WWD 1 Purpose and Overview 2 1 Purpose and Overview This User Manual for the ETACE Virtual Appliance describes how to conduct economic analy ses with the Eurace Unibi model using the programmes available on the VA and its respective tools with regard to the general workflow of the Flexible Large scale Agent based Modelling Environment FLAME The virtual appliance has been created at ETACE the Chair for Economic Theory and Computational Economics at Bielefeld University The intention behind this software collection is to make every step related to the initialization execution modification and analysis of the Eurace Unibi agent based simulation model as easy as possible We hence address here the issue of reproducibility of simulation based research Stodden 2010 2 Programmes Documentation and the Eurace Unibi Model This software package is based on the SliTaz dist
6. specified Choose Agent Choose Variable Instance Selected Time Series Firm output mean_specific_ skills Eurostat wage offer technology price price 1 Add gt gt labour _costs_ per_unit lt lt Remove Discard e Double click on the agent type from which the time series should be added e After choosing the agent you can now select the time series from the list of available variables e Confirm the settings by OK A The time series are plotted by using the same scale of the y axis Defining default settings Io define default settings use the Default Set z tings item of the menu Settings One Tmax 28000 _ can set default values for the minimum OK _ Discard iteration and the maximum iteration 4 How to use the Simulation GUI for customized Simulation Experiments 18 4 4 4 How to plot Agent Distributions Time Series Multiple Time Series The Agent Distribution tab can be used to analyze the distribution of agent variables in the population The distribution of a variable is plotted as a box plot and as a histogram for a point in time or for a discrete sequence that has to be specified by the user By default the distribution is plotted for the last iteration of the simulation To analyze the distribution of an agent variable switch to the corresponding agent There is a table listing the variables currently used for plotti
7. 4 How to use the Simulation GUI for customized Simulation Experiments 19 Defining filters same as for adding new variables for time series plotting see Page 15 Specifying time snapshots for distri butions Click the Iterations item in the Settings menu which opens a dia 5000 log window in which the iterations can ae E a be selected for which distributions are plotted One can enter a sequence of it erations by entering comma separated values Define default setting profile Click the Default Setting item in the Set Default Settings Mean Default Filter 1 No Filter tings menu which opens a dialog win dow in which the default settings can be defined The default settings are those settings that are automatically chosen when defining new time series Default Filter 2 INo Filter Tmin 1 Tmax 28000 4 4 5 Plotting Output The GUI launches the plotting process automatically after the simulations have finished During the plotting process the R jobs generate plots and write these plots to files in a specific sub folder structure These sub folders can be found in the Plots folder located in the workspace The Plots folder contains three sub folders e Boxplot which contains the plots showing the 7 E level folder op_level_tolder agent distribution as box plots gt its gt lcs e Histogram which contains the plots showing the agent distribution as histogr
8. Setting tab Defining Time Series To generate a time series of a memory variable of an agent the user has to specify settings Besides the agent type and the name of the variable those settings include a method which is applied to aggregate the data and filters used to draw a sub sample of the population featuring certain characteristics The agent tab contains a table listing the currently selected time series The table also provides an interface to edit the time series settings 4 How to use the Simulation GUI for customized Simulation Experiments 14 Time Series for Agent Firm Variable Name _ Method Filterl Filter 2 Further setti output output Mean v NoFilter_ No Filter e technology technology Mean v NoFiler NoFilter mean spedific mean specitic mean e No Filter_ w No riter Lz wage offer wage offer mean No Fiter _ w No riter price price mean e No Fiter No Fiter ____ labour costs labour costs_ Mean No Filter _ w No Filter z e Variable The name of the variable used in the model e Name A customized name of the time series Change the name by editing the corre sponding field of the table This name is used as the file name when generating the plots and will appear on the y axis of the plot A Hit return after editing a time series name e Method Pull down menu to define the function applied to aggregate the time series One can use M
9. agents to store at which frequency every iteration every month 3 Running a set of batch runs 4 Launching the simulation 5 After the simulation has finished and all data has been produced the final stage is the data analysis stage For this we again make use of the Simulation amp Analysis GUI 4 1 Getting Started Define the Work Space and Enter basic Settings The GUI works with a workspace environment This means there is a folder in the file system of the VA that is defined by the user as the location where all files related to a simulation experiment are located How to set the Workspace There are two possibilities to set the current workspace 1 Set the workspace immediately after launching the GUI a file browser opens after launch ing the GUI which can be used to load an existing workspace or to create a new workspace e Load an existing workspace use the file browser to browse to the top level folder of the workspace Click Open to load the settings The GUI reads the saved settings of the workspace and opens the main dialog window e Create a new workspace use the file browser to browse to a location in the file system where the new workspace should be created Use the new folder button to create and name the workspace In the following the user has to set some basic settings manually Use the file browsers to set the path to the model xml file eurace_model xml the main executable main the i
10. ams nF __ Plots bP Boxplot b Histogram __ Time_Series e Time_Series which contains the single and mul tiple time series The Time_Series folder has three sub folders 4 How to use the Simulation GUI for customized Simulation Experiments 20 e Single This folder contains for each single run of each batch all single time series plots and mul tiple time series plots e Batch Summarizes the time series plots of the g Pts batch analysis The directory name of the batch p Boxplot T BB Histogram sub folders indicate for which parameter value w I Time_Series the batch plots have been generated There are v i Batch three sub folders Ay __ batch_0 0 TE Full Contains plots that show all time se Tidi ries of the batch runs p Multiple kp Qpbatch_11 Mean Contains plots which depict the av D H Parameter erage time series of a batch in a single v GB Single plot The time series are averaged across v jj batch_o o gt o run D gt Gunz Multiple Contains the multiple time series i b un the time series included in the plots are av p gt uns eraged across the batch runs i gt runs p batch_1 1 the batch runs e Parameter Contains time series plots that shows for each of the considered parameter values the average time series of the batch in a common graph This can be used to compare the effects of a variation of the parameter of the experiment The f
11. at print_debug ch print_debug afm print_debug afm_ch print_ debug expl iprint_debug file _exp1 print_debug file _exp2 policy exp1 print_ debug gov print_debug household print_debug credit print_debug consumption A Hit return after entering a new value e Confirm the changes SISO SO oO Oo olSolol Solo ooo So olololol oo oun print_debug finman print_debug production print_debug d h di 7 print_debug_market_research an CLOSE t e la og Confirm policy_exp_energy_shock policy_exp_stabilization_subsidy policy_exp_stabilization_tax window const_bankruptcy_idle_period 240 days_per_month 20 xmi dond 0 e Discard the changes individual_learning 0 const_wage_wealth_ratio 0 200000 const_income tax_rate 0 050000 and close the dialog Discard gamma _const 9 000000 gamma _ quality 1 000000 depreciation_rate 0 010000 d discont_rate 0 020000 window mark_up 0 200000 t t lambda 1 000000 R ll h target_savings rate 0 100000 e ese a C anges O carrol_consumption_parameter 0 010000 gquantil_normal_distribution 0 524400 h d f f firm_planning horizon 10 t e au t Va ues O adaption _delivery volume 0 100000 inv inertia 10 000000 h d gamma logit vintage choice 30 000000 t e exp eriment Save Confirm Discard Reset nN the O xml file 4 2 2 Experiment Mode Select whet
12. ean Median Standard deviation Percentage standard deviation Sum Minimum Maximum Weighted averages this method can only be selected after specifying a weighting factor in the Settings menu e Filter 1 and Filter 2 The user can apply two filters on the time series The filters can be defined in the Settings menu e Further Settings Clicking the button opens a dialog window to define tmin and tmax Defines the limits to the x axis i e the range in terms of iterations within which the time series is plotted Lower Bound and Upper Bound Customized limits of the y axis A The same memory variable can be used for different time series Adding new Time Series There are two possibilities to add new time series Which one to use depends on whether or not the memory variable from which a time series should be plotted has already been selected for plotting time series 4 How to use the Simulation GUI for customized Simulation Experiments 15 e Ifthe variable has not yet been used id mean_specific_skills for plotting time series Click the Ad region i output d Remove Time Series item from at of manih toaa page aer the Time Series menu bar This banka opens a dialog window which can be mean wage used to select variables from the vari needed capita stock able list of the considered agent Mark aaua can prio adda one or more variables and click Add planned output Ls Remove to s
13. ed parameter being varied Do not compress Firm fi _t 3 Data management which simulation O Compress Databases and remove Original D Compress Databases and keep Original aro a 7 output will be st ored for the post O Remove decompressed Database Centra c simulation analysis O Decompress Database Write all agent variables to Database 4 2 1 General Setup Launch simulations and plotting or plotting only These radio buttons define whether the GUI launches both the simulations and the plotting process or the Run Simulations Do not run Simulations plotting process only This feature can be helpful if the simulation data has already been generated and the user wants to change only plotting settings Set number of batch runs 4 How to use the Simulation GUI for customized Simulation Experiments T Enter the number of batch runs For each parameter value of the experiment the GUI launches the corre Number of batch Runs 10 sponding number of simulations The runs only differ by the random seed used A Entering new values in any text field must always be confirmed by hitting enter Note on random seeds The random seed is a parameter model constant in the model xml file called RND_SEED that needs to be initialized or else defaults to zero This allows us to include the value of the random seed in the 0 xml file and then use it as an in
14. elect these variables as time series proc costs One can also remove time series which chnoogial aes should not be plotted any more by ital technology marking them and clicking Remove oo quality Clicking Apply Changes confirms the ws changes and closes the dialog window ay eer Ga A The list of variables can be ordered alphabetically by clicking the right mouse button e Ifthe variable has already been used for plotting time series Mark the time series in the table and click the Add Modified Time Series button The GUI adds a new line to the table con taining a new time series of this vari able with the default settings The name of the new time series is the name of the variable plus a number as each time series need a unique name Defining filters Filters can be de fined by clicking the Filter item in the Settings menu which opens a dialog window for editing filters Filters can id region idi 2 be used to define sub populations of an AN a agent type by filtering out the agents oe ee toact add that do not have the requirement de payment account balan fined by the filter By clicking on one a variable from the list one selects the fil E ter variable Another window pops up to enter the filter options filter value and filter method The filter methods are 4 How to use the Simulation GUI for customized Simulatio
15. em s BIOS to run in multi core mode normally under menu CPU in BIOS e In the settings of the virtual machine itself i e in the client fi Oracle s Virtu alBox but not inside this virtual appliance you can change the number of assigned CPU cores and RAM memory To get root access in a terminal type su with password root The Super User is likewise root with password root All the relevant files have been placed on the Desktop home eurace 4 How to use the Simulation GUI for customized Simulation Experiments 4 Desktop whereas additional libraries Libmboard R Python GSL have been installed directly into the system 4 How to use the Simulation GUI for customized Simulation Experiments The SimulationGUI has been developed to have an all in one user interface for setting up running and analyzing simulations of models implemented in FLAME The general design of the GUI acknowledges that setting up and analyzing simulation ex periments are two separated tasks the first done before the simulation and the second after the simulations have finished These tasks are also separated in the GUI which is achieved by providing two separated tabs one for the simulation settings and another for the post processing of the data The Simulation GUI helps you to perform the following steps Design of the Computational Experiment 1 Setting the environment constants model parameters 2 Selection of data storage options
16. ent population with the specified phase and periodicity A Hit return after entering a new value Record all agent variables These radio buttons are important for the space re quirement of the VA e If Yes is selected then the full memory of each selected agent is written to disk As one typi cally needs only a small subset of agent s mem ory variables for the post simulation analysis Write all agent variables to Database Yes No this setting can imply a waste of hard disk space especially if running large simulations e If No is selected then only a small subset of agents memory variables are recorded The se lection of memory variables to be recorded is based on the selection for the plotting settings see Section 4 4 If choosing option No only the variables that have been selected in the plotting set tings are recorded Be aware of the fact that non recorded data can only be recovered by re running the simulation It is highly recommended to choose the variables in the plotting settings carefully before starting larger simulation experiments 4 How to use the Simulation GUI for customized Simulation Experiments 11 4 3 How to build and Run experiments time_series_data txt scatter_data txt correlation_distribution_data txt STATUS growth_rate_data txt multiple_time_series_data txt variables txt heat_maps_data txt boxplot_data txt cross_correlation_function_data txt bandpass filter_data txt
17. ents In principle FLAME also allows agents of a specific type to send messages to agents of the same type 5 4 Population GUI Agent Population Instantiation After the entire model design stage is done the next stage is to initialize our agent population This initialization is done using the PopGUI This stage consists of setting the size of the agent population the number of regions and the subpopulation of agents in each region We can initialize all model constants and agent memory variables For each agent memory variable we can define relationships on the initialization values of other memory variables of the same agent or of other agents or on the values of model constants It is possible to validate the complete set of relationships before instantiating the population At the end of this stage a complete population description file called 0 pop file has been generated with all the interdependencies between the agents initial values resolved From this the input file called 0 xml file is generated In the case you want to use FLAME to develop a new model a tutorial and further explana tory materials can be found in the Documentation folder To explain this in detail would go beyond the scope of this user manual 6 Licensing The ETACE Virtual Appliance is made available under the GNU General Public License This does not affect any contributions to the Appliance by others which are included with permission of
18. fer here to the SliTaz documentation e Once you have created a new model xml or 0 xml file use the xparser to parse and compile the model After double clicking the desktop shortcut choose the model xml file usually called eurace_model xml After x parsing the model you will be prompted to press ENTER to compile the model For advanced users the GUI version of the xparser does not provide any additional options If you require these you will have to run the xparser from the command line home eurace Desktop Models xparser gsl xparser Options include p parallel code f final production mode 5 Flame Modeling Environment 22 Virtual Appliance VA Step 1 Development FLAME Editor gt Model xml Codelite Eclipse Editor gt Agentfunction s c Step 2 Parsing Compilation 1 Xparser gt Simulationcode c 2 GNU C Compiler make gt main Agentfunction c Liomboard Step 3 Initial Data Generation Execution PopulationGUl gt 0 xml file SimulationGUl gt Generation txt Data Generation 1 N xml files Step 4 Processing 1 Data Extraction gt XML gt SQL 2 Data Transformation 3 Data Visualization gt png pdf Fig 1 Overview of the Flame Workflow and Programmes in the ETACE Virtual Appliance 6 Licensing 23 5 3 Flame Editor Model Design Stage The model design stage starts with setting up the general model h
19. he workspaces The settings which can be imported exported are e Plotting Settings Exporting and importing the settings related to the data analysis This is to use different plotting profiles Exporting Use the file browser to define a location to which the plotting settings are exported The GUI writes a xml file with the currently used plotting settings The name of the file has to be specified by the user Importing Use the file browser to browse to the export location Select and import the xml file in which the plotting settings have been exported e Parameter Settings Export and import parameter settings in order to switch between different parametrizations of the model Exporting Use the file browser to define a location to which the parameter settings are exported The GUI writes a xml file with currently used parametrization of the model The name of the file has to be specified by the user Importing Use the file browser to browse to the export location Select and import the xml file in which the parameter settings have been exported Setting the required file paths manually When creating a new workspace from scratch the user is automatically asked to set the correct path to several file resources required by the GUI However these paths can also be set manually It should be noted that these path are mandatory and have to be set correctly before setting up an experiment Use the
20. her or not a model parameter is varied These radio buttons can be used to switch between two experiment modes Run only one Batch 1 Run only one Batch run a batch only with the current parameter setup Parameter Variation with one Parameter 2 Parameter Variation with one Parameter a se lected model parameter is varied among specified values Select the model parameter for the experiment 4 How to use the Simulation GUI for customized Simulation Experiments Edit Parameter 1 probability tec 1 1 gov_policy_switch_human_capital_improvement a subsidy_redistribution tech_policy_effectiveness switch robability tech_policy effectiveness policy_exp_best_technology subsidy subsidy redistribution _parameter 4 2 3 Data management Compress or decompress the databases Do not compress OK Discard O Compress Databases and remove Original O Compress Databases and keep Original O Remove decompressed Database O Decompress Database Data recording of agents Clicking Edit Parameter 1 opens a dialog window in which the user can select the parameter to be varied in the experiment as well as can define the parameter values used in the experiment This button is only active if the radio button Parameter Variation with one Parameter is selected e Scroll through the list of parameters and choose the parameter for the experiment by clicking on that parameter
21. ierarchy The entire model can be subdivided into several modules each having an internal structure that adheres to the XML definition Each module will contain a well defined description of the agents as X Machines In fact each module itself is also an X Machine and the model as a whole can also again be construed as X Machine c f Holcombe 1988 Balanescu et al 1999 In the next stage we define the agent types their memory variables functions and the activation structure of the functions scheduling of functions can be either time or event based or both At this stage we also define message definitions and set the input and output messages for the functions In addition the environment constants fixed model parameters can be defined The complete XML structure can be tested for consistency before writing any actual code by using the Flame Editor GUI developed by Simon Coakley from Shefheld University After Xparsing the full model the model design stage ends and the hierarchical structure can be inspected through a birds eye view of the model in a stategraph an example of which is given in Figure 2 Such a stategraph shows for every agent its states and the transition functions between states It also shows the branching of agent activities depending on time conditions monthly or yearly activation or on event based conditions a memory variable of an agent In addition the stategraph also shows the flow of information between ag
22. imulation data as output to file The progress of the simulation is indi cated by a progress bar which automatically pops up when the simulation is being started For writing the simulation data the GUI creates a folder with a spe cific hierarchy in the workspace with its as top level folder Each simulation run is written in a specific sub folder of the hierarchy whose path relative to its can be used to identify the particular run The simulation data is first written into iteration specific xml files e g 20 xml 40 xml etc In a second step these data files are translated into SQL databases and finally deleted Closing terminals or the GUI ends the sim ulation Terminals or the GUI should only be closed if the user wants to terminate the simulation The data processing The SQL databases are read by R in order to carry out a user defined data analysis The results of the data analysis are written to files that are stored in a folder system below the Plots folder which is created after the simulations have finished The log folder which is also created during the data analysis contains the console output of the R scripts and can be used to read error messages in case the figures have not been created correctly 4 How to use the Simulation GUI for customized Simulation Experiments 12 How to run multiple experiments inarow There is also a possibility to run more than one ready to run experiment automatically witho
23. mic and biological models It uses the markup language XMML X Machine Modeling Language for the declaration of function and variables and template files in C to generate the final C code of the model All model functions are written in plain C code a small subset of C while the scheduling of agents functions and messages is coded in XML with the help of an easy to use FLAME Editor The XML model file is then parsed by the Xparser producing C code for the simulator This C code is then compiled together with the user provided C code for the agent functions which generates the simulation executable Libmboard Message Board Library provides facilities for message passing 5 2 Flame Model Workflow A detailed workflow can be found below For producing new models quickly we note that e To view a model and its agents functions messages etc use the FLAME Editor and open the model xml file e To generate initial data in an 0 xml file use the Population GUI Also use this to view or edit pre existing pop files population description files These files can be found in the same its folder as the default 0 xml file e In order to keep this image lightweight we do not provide a front end for C programming To view and edit the code you can use vi nano or leafpad all available from the terminal If required additional programmes can be installed using the SliTaz package manager tazpkg For further detail on that we re
24. n Experiments 16 e Ignore agents with variable filter value Ignore agents with variable gt filter value e Ignore agents with variable lt filter value H Ignore agents with variable filter value By clicking Apply the settings are confirmed and the new filter can be selected from the pull down menu Filter 1 and Filter 2 in the time series table Define a default settings profile Click Default Settings in the Set Default Settings Mean Default Filter 1 No Filter tings menue which opens a dialog win dow in which the default settings can be defined The default settings are those settings that are automatically chosen when defining new time series Default Filter 2 No Filter Tmin 1 Tmax 28000 Define weighting factor for weighted averages Click Select Weighting Factors for Means in the Settings menu which opens a dialog window in which weighting factors can be defined to compute weighted averages Select asa i TTT variables from the list to have the pos day of month too sibility to compute averages over the neo ae payment_account_balanc population weighted by this variables ae The additional aggregation methods ie are can then be selected as additional options in the pull down menu of the time series table A possible applica tion is a price index for which an av erage price weighted by fi
25. ng and which filters are applied i e the properties of the considered sample of the population Plotting Histograms and Boxplots for Agent Firm Variable Name Filter 1 Filter 2 unit_costs unit_costs No Filter No Filter v output outpt No Filter No Filter z Adding new Distributions There are two possibilities to add a new distribution Which one to use depends on whether or not the memory variable for which a distribution should be plotted has already been selected for plotting distributions e If the variable has not yet been reglon used Click the Add Remove vari SE day of month toad ables item from the Agent Distribu anki tion menu bar This opens a dialog window which can be used to select variables from the variable list of the considered agent Mark one or more variables and click Add One can also remove distributions from the list by marking them and clicking Remove Clicking Apply Changes confirms the changes and closes the dialog window Apply Changes Cancel A The list of variables can be ordered alphabetically by clicking the right mouse button e If the variable has already been used Mark the distribution in the table and click the Modify variable button The GUI adds a new line to the table featuring the default settings The name of the new distribution is the name of the variable plus a number as each distribution needs a unique iden tifier
26. nitial start state 0 xml The top level folder of the R scripts The default is 4 How to use the Simulation GUI for customized Simulation Experiments 5 home eurace Desktop src JavaGUI src Data_Analysis_GUI_Serial The executable of the xparser The default is home eurace Desktop src xparser 0 17 1 xparser 2 Use the Experiment menu of the main menu bar and click Load to open an existing workspace Use the file browser to browse to the top level folder of the workspace and load the workspace by clicking Open To create a new workspace click New Use the file browser to browse to the location where the new workspace should be created The new workspace is initialized with the current settings Save the current state of the workspace The current state of the workspace can be saved by using the Save item of the Experiment menu The previously saved settings are overwritten Alternatively one can create a new workspace by using the Save as item of the Experiment menu In this case one gets a new branch of the workspace without overwriting the old settings Importing and Exporting special configurations There can be the situation that the user wants to flexibly use different settings as e g plotting settings without switching between workspaces For this case the user can use the import and export feature of the GUI Use the Import Export item of the Experiment menu to export and import settings without branching t
27. older structure for distributions is the same for boxplots and histograms and is almost the same as used for time series Plots v __ Boxplot i v _ Batch gt J batch oo e Batch Summarizes the distribution plots of the Sb Gp batch batch analysis The directory name of the batch ie ere sub folders indicate for which parameter value o i ar ee the batch plots have been generated o alam This older convas ISA ution Lb pnuns plots showing in one plot the distributions for gt D nA each batch of the considered range of parameter b nuns values p Me batch_1 1 gt Jj Histogram e Single This folder contains distributions for rb f Eater each single run of the batches p D Parameter __ Single gt Time_Series 5 Flame Modeling Environment 21 5 Flame Modeling Environment The Eurace Unibi model is implemented in the Flame development framework We provide the full source code for the EuraceQ Unibi 1 0 model If you want to get a deeper understanding of this code make changes i e implement a new policy rule or experiment with different initial populations continue reading A Flame model development cycle goes through several stages as illustrated in Fig 1 5 1 What is FLAME In a nutshell FLAME is a program generator It generates a simulation executable from C and XML files FLAME is a domain specific language programmed in C that provides facilities for higher level programming of econo
28. ontains the licenses of the ETACE Virtual Appliance as well as manuals for GUIs and applications used in the VA Furthermore it contains a description of the Eu raceQUnibi model together with several research papers based on eae eae the model e src this folder contains source files of applications e Models this folder hosts source code and executables of different versions of the Eurace Unibi model e exper this is an empty folder for storing new simulation experi ments of the user e Preconfigured Experiments this folder contains ready to run simu ing f t Pea rseeririeuirse ation setups for exact paper replication Experiments The launchers can be used to launch different GUIs e PopulationGUl to create an initial state as input of a simulation e AgentGUI to change the model structure e xparserGUI to parse and compile the model e SimulationGUI for setting up and starting simulations as well as post processing simulation data SECO Only the SimulationGUI is required for replicating the experiments of the paper Performance and the Linux system The virtual appliance is set up with a minimum performance configuration You can increase the performance of the VA especially by increasing the number of processors and the size of the allocated memory However the performance is limited by the configuration of the host system To change these A e Enable hardware virtualization support in your own syst
29. ot run Simulations 4 4 1 Basic plotting settings At the bottom of the plotting tab there are elements that can be used to set some basic plotting settings e The check boxes Single Run Analysis Batch Run Analysis and Parameter TEE a 0 Ti Analysis can be used to opt whether the Batch Run Analysis e post simulation analysis should include an Parameter Analysis Site Type of Pots analysis of single runs of batch runs and a detailed analysis based on the varied pa rameters Transition Phase This allows to cut off an initial transient of the time series A This is not the same as the transient phase used to create the initial state file 0 xml This transient phase is not stored e Add legends Specifies whether legends are added to plots e Colored Plots Specifies whether the lines of the plots are colored or black File Type of Plots Select the file type at the moment only the pdf format 4 4 2 How to plot Single Time Series In order to plot single time series one has to switch to the Time Series tab of the Plotting Settings The tab Time Series itself contains a set of tabs each tab for an agent type of the model To generate time series for different agent types one can simply switch between the agent tabs In order to generate time series for an agent type one has to make sure that this par A ticular agent type has been selected in the data recording table on the Simulation
30. put into the simu lation This way for all runs we use a different random seed and still have full reproduceability of the simulation results because the value of the seed is stored in the initialization files Set number of iterations l l l l Enter the number of iterations Each simulation run is executed for the indicated number of iterations In the Eurace Unibi model one iteration corresponds to one day Considering only working days 5 iterations constitute a week 20 iterations a month and 240 it erations a year A good starting point for a standard experiment is 5000 iterations which corresponds to 20 years Number of Iterations 28000 Set number of process threads This pull down menu can be used to set the number Number of Processes h x of simulation processes This is helpful to reduce the simulation wall time of an experiment A It is recommended to execute simulations in multiple process threads but to limit the number of threads to the number of cores with which the VA is configured Change the parametrization of the model 4 How to use the Simulation GUI for customized Simulation Experiments 8 This button opens a new dialog window which can be used to change the value of any model parameter Change Parameter Setup Parameter total_regions id debug probe print_log print_debug bankruptcy e New parameter values can be entered in the table print_debug eurost
31. ribution of free software that includes the Linux kernel The coresponding documentation is included in the software Programs provided in the Virtual Appliance including their dependencies Xparser GUI Parser for FLAME models GNU GCC compiler C compiler for model framework code Flame Editor Generate model xml file XML description of model Population GUI Generates initialization files population description Simulation GUI Settings for simulation experiments and data analysis Apart from standard dependencies such as GNU GCC the relevant documentation to these programmes can be found in the Documentation folder on the Desktop The following versions of the Eurace Unibi model are included and can be found on the Models folder on the Desktop Dawid et al 2011 full source code of EuraceQ Unibi 1 0 Dawid and Gemkow 2013 main model xml 0 pop amp 0 xml for replication Dawid et al 2014b full source code of the model used for the paper Dawid et al 2014a full source code of the model used for the paper For the following papers a pre configured ready to run experiment can be found in the Preconfigured Experiments folder Dawid et al 2014a XXX 3 Quick Starter Guide The Desktop After launching you should see the desktop of the VA The desktop contains folders and program launchers The folders are l http www flame ac uk 3 Quick Starter Guide 3 e Documentation this folder c
32. rms sales is computed 4 4 3 How to plot Multiple Time Series Multiple Time Series Agent Distributions All time series defined on the Time Series tab are plotted as a single time series graph in a single plot If you want to plot more than one of these time series together in one plot one you use the Multiple Time Series tab to define those multiple time series Time Series A Only time series that have been selected for plotting as single time series can be plotted in a multi plot The Multiple Time Series tab contains a table listing the currently used multiple time series and the corresponding settings 4 How to use the Simulation GUI for customized Simulation Experiments 17 Multiple Tmin Tmax Lower Limit Upper Limit mts tec 1 28000 List o e Multiple Time Series Name The name of the time series the name is a combination of the prefix mts and the names of the included time series This name is used as file name when generating the plots e Tmin and Tmax as minimum and maximum iteration numbers to appear in the plot e Components This shows a list of the time series e Lower Limit and Upper Limit Customized limits for the y axis Add Multi Time Series In order to create a new multi time series or to edit an existing one you can select the Multiple Time Series menu and click Add Time Series or Edit Time Series This opens a dialog window in which a multiple time series can be
33. the authors and whose rights remain non infringed upon Use of the source code of the EuraceQ Unibi model is guided by the End User License Agree ment EULA that can be found in the Documentation folder of the Software Package If you cannot find this file in your copy of the Software Package or if you claim any further rights on some parts of this work please contact ETACE wiwi uni bielefeld de 3 The XMML of FLAME has been defined as a hierarchical DTD Document Type Definition This means that FLAME accepts a nested model definition consisting of sub models which again may consist of sub sub models and so on and so forth This is possible due to the fact that a collection of Communicating Stream X Machines can again be considered an CSX Machine REFERENCES 24 fox_start_state rabbit_start_state layer O Fox_send_location Rabbit_send_location 01 01 fox_location rabbit_location rabbit_location a a layer 1 GlobalAgent_count_agents Fox_chase_all_rabbits fox_location GlobalAgent_end_state 02 rabbit_eaten layer 2 Fox_starvation Rabbit_dodge_foxes Fig 2 State graph of a simple predator prey model with foxes and rabbits A full list of all the contributors to this Appliance or the included software as well as their respective licenses can likewise be found in the Documentation folder References Balanescu T Cowling A J Georgescu H Gheorghe M Holcombe M and Vertan C 1999 Communica
34. ting Stream X machines systems are no more than X machines In Twelfth International Symposium on Fundamentals of Computation Theory FCT 99 Iasi 5 494 507 Dawid H and Gemkow S 2013 How do social networks contribute to wage inequality insights from an agent based analysis Industrial and Corporate Change Dawid H Gemkow S Harting P van der Hoog S and Neugart M 2011 The eu raceQunibi model An agent based macroeconomic model for economic policy analysis Avail able from www wiwi uni bielefeld de vpl1 projects eurace eurace unibi html Dawid H Harting P and Neugart M 2014a Cohesion policy and inequality dynamics Insights from a heterogeneous agents macroeconomic model DFG Research Center SFB 882 From Heterogeneities to Inequalities 34 Dawid H Harting P and Neugart M 2014b Economic convergence policy implications from a heterogeneous agent model Journal of Economic Dynamics and Control 44 54 80 Holcombe M 1988 X machines as a basis for dynamic system specification Software Engi neering Journal 3 2 69 76 REFERENCES 25 Stodden V C 2010 Reproducible research Addressing the need for data and code sharing in computational science Computing in Science amp Engineering 12 5 8 12
35. ut launching each of them separately Use the Run Batch item of the Experiment menu This opens a dialog window for selecting ready to run experiments that can be executed without separate launching e Click Add Experiment in the menu of the dialog window to add an experiment to the list Use the file browser to load the experiment xml file of the corre sponding experiment which is located in the workspace folder of the experiment Experiment Batch Add Experiment Remove Experiment Exit Path Parameter 1 Values Parameter 2 Values shome eurace De probability tech_ Values fv Run Batch e If the list of experiments is complete click Run Batch The GUI starts to run the first experiment on the list and continues with the next after finishing the previous experiment 4 4 How to set up the Post Simulation Analysis Simulation Settings Plotting Settings The GUI provides basically ways of analyzing the simulation data of an experiment The one is to consider aggregated time series and the other distributions of agents memory variables In order to configure the post simulation analysis the user selects the Plotting Settings tab of the GUI The user can switch between three tabs Simulation Settings Plotting Settings Time Series Multiple Time Series Agent Distributions Time Series Settings Firm Household Mall IGFirm
Download Pdf Manuals
Related Search