Home

Microsimulation model user guide

1. 10 V Randomise Figure 39 Manualling identifying the zone The link is now ready to be saved All of the relationships are defined and the zone identification field has been determined To save the link right click in the link table area to reveal the context menu shown in Figure 41 select Save link The link table area resets the Save configuration and 34 Run buttons become enabled again and the context menu in the population table area is again accessible You will also notice that the name of the link table has been added to the drop down box at the bottom of the screen Table for zone totals we will come to this later The Remove row and Add row options from the context menu allow for the correction of any mistakes made during the specification of relationships between the sample population table and the constraint It is worth ij Microsimulation Population Table Population Optimisation settings set the maximum number of steps in the optimisation algorithm 100 set the maximum number of improvement attempts before continuing Linked field _ 1L 100 set the number of improvements to the optimisation before continuing 10 set the annealing reduction factor recomended 0 9 H os Pop id field Remove row Save link lt lt output table group name gt gt Output to lt lt output data source gt gt Da Save
2. National Centre for Research Methods Working Paper 06 13 Microsimulation model user guide Flexible Modelling Framework Kirk Harland TALISMAN node University of Leeds NCRM National Centre for Research Methods UNIVERSITY OF LEEDS Microsimulation model user guide Flexible Modelling Framework Kirk Harland Version 1 0 September 2013 School of Geography University of Leeds Leeds LS2 9JT United Kingdom This Working Paper is an online publication and may be revised Talisman Multi Agent Systems si T and Simulation Contents 1 Introduction to static spatial microsimulation 1 1 Idealised example 2 Installing the software 3 Starting the software 3 1 Navigating the screen 3 2 Application menu 3 2 1 Shutting down 3 2 2 Setting the data directory 4 Data Sources 4 1 Registering and unregistering a data source 4 2 Registering and unregistering files 4 3 Loading data and opening tables 4 4 What happens if file names or directories change between loads 4 4 1 Change in data source directory 4 4 2 Change in file name 4 4 3 Change in file structure 4 4 4 How Null is treated 5 Windows and Processes 5 1 The windows tab 5 2 The processes tab 6 Microsimulation 6 1 What is the technique and what does it do 6 2 Data format 6 2 1 Structure of the sample population 6 2 2 Structure of the constraints and evaluation tables 6 2 3 How the tables relate 6 3 How the algorithm works
3. 6 5 1 Creating links between the sample population and constraint tables To begin creating links between the sample population and the constraint tables first locate the sample population table in the data sources left click to highlight it and while keeping the left mouse button pressed drag it to the population area of the microsimulation screen and release the mouse button The fields from the population table will now be displayed in the Fields column in the population table area of the screen and the label above will display the name of the table in this ij Microsimulation Population Table Population set the maximum number of steps in the optimisation algorithm 100 set the maximum number of improvement attempts before continuing 100 Age_Gender set the number of improvements to the optimisation before continuing Link Table 10 set the annealing reduction factor recomended 0 9 p Output to lt lt output data source gt gt lt lt output table group name gt gt Save configuration lt lt model name gt gt Table for zone totals 10 IV Randomise Figure 35 Population table displayed example Figure 35 Population Table Population because the table used here is called Population We now need to tell the model which field in the sample population table is the unique identifier Locate the field in this example it is P_ID First left click
4. Application Microsimulation j Reister Flat Files Msim Age cs imi Text Qualifier Age by Ethnicity csv Ethnicity csv Gender csy fe Gender_by_Aqe csy e Population cs Sex _by_Ethnicity cs f none Headers in First row Register Figure 14 Register tables window A list of files available to be registered in the data source is shown in the area highlighted by the red box in Figure 14 Selecting a file from the list will load the top 10 rows of the data and display them in the sample are shown in Figure 15 below In this case the file Age csv has been selected and the sample data is shown in the area highlighted by the red box 15 ig Flexible Modelling Framework version 1 0 Beta Application Microsimulation ig Register Flat Files Msim Age_by_ Ethnicity csv Ethnicity cs v Gender cs Gender_by_Age cs Population cs Sex_by_Ethnicity csv Fido medi Field2 Field3 reda meds Zone Code fotoa Sto Bto fi0toi4 f i fia ooparaceol heo jao eo wo po k o bo eo wo oopaFaoooa 26 0 pao po so oarso T povaraooos 90 0 Bo 7 0 14 0 a ain Register Figure 15 Register flat file window with file selected Notice that the field names have been automatically added as Field 0 Field 1 etc If your data has field names in the first row select the Headers in first row check box as shown in Figure 16 below to use these headers fg Flexible Modellin
5. Eval UG Test _Gender_by_ Age F Eval UG Test_Age_by Ethnicity Gl Eval UG Test_Sex_by Ethnicity Figure 67 Example summary tables created Once the summary tables have been created the evaluation can be run to generate the fit statistics and fit summary tables The fit statistics are calculated by comparing the observed values in the constraint and evaluation tables to those created from the synthesised population in the summary tables Three types of fit table are created see Figure 68 1 An overall summary fit table is created with the name lt lt output prefix gt gt _gof in the example here Eval UG_Test_gof The table will contain a field called VariableName which will have a record with the name of each table linked to the sample population table It will also contain a field for each fit statistic chosen and an overall value for the statistic for each constraint evaluation table in the relevant row A summary fit table by zone with the name lt lt output prefix gt gt _zones_gof in the example here Eval UG Test_zones gof The table will contain a field with the full list of zone identifiers followed by a field for each constraint evaluation table fit statistic combination following the naming convention of lt lt table name gt gt _ lt lt statistic abbreviation gt gt This table provides a summary of the level of fit by zone The final level of fit table if for each constraint evaluation table and follow the n
6. Register tables window Register flat file window with file selected Register flat file headers delimiter and text qualifiers Data type context menu Registered table appears in data sources tab Expanded table showing field types Files already registered no longer show in file selection area Un registering a table Un register table confimation dialog Data cached in sample tables Table context menu Table window opened in the action area of the screen Invalid data source Interaction between windows tab and open windows in the action area 10 11 11 12 12 13 13 14 14 14 15 15 16 16 17 17 18 19 20 20 21 21 22 22 24 Figure 28 Figure 29 Figure 30 Figure 31 Figure 32 Figure 33 Figure 34 Figure 35 Figure 36 Figure 37 Figure 38 Progress bars for processes in the processes tab Example sample population Constraint table example structure Relationship between the sample population and constraint table The Simulated Annealing algorithm Microsimulation configuration screen Relationship of Simulated Annealing algorithm and configuration controls Population table displayed Selecting the unique identifier Pop id field displayed in action column Link added to population table field Figure 39 Manualling identifying the zone identification field Figure 40 Figure 41 Figure 42 Figure 43 Figure 44 Figure 45 Figure 46 Figure 47 Figure
7. White_British 0 _to_15 Male People 0_to_4 _ White_British Male_ White_British O_to_15_Male_People O_to_15 Male People S_to_15_White_British IS_to_15_White_British Male_ White_British Male_ White_British O_to_15 Male People IS_to_15_White_British Male_White_British white _ pa O_to_15 Male People S_to_15_White_British Male_White_British Molol ath Eilan eS White_British 0 _to_15 Male_People IS_to_15_White_British Male_ White_British White _ British O_to_15 Male People S_to_15_White_British Male_White_British tt White_British O_to_15 Male People S_to_15_White_British Male_ White_British e e l Ea m White_British O_to_15 Male People S_to_15_White_British Male _white_British white _British 0_to_15_Male_People 5_to_15_white_British Male_white_British 5_to_15_white_British Male_White_British White _British 0_to_15_Male_People white_British 5 _to_15_White_British Male_White_British White_British 16 to 19 hae People 116_to_29_White_ British Male_White_British Mj table Gender_by_Age Zone Code f 0_to_15_Male alia 0_to_15_Female Pempe 4 16_to_19_Male_People 16_to_19_Female_People 20_to_24_ OF ODO MO VM aa rn nn oe ain Figure 31 Relationship between the sample population an
8. attributes from both aggregate and survey data This allows estimates of demand or prevalence to be made for small geographical areas using attributes not collected at that spatial resolution a powerful tool for policy formation and local or regional planning Alternatively the synthetic population produced by a static spatial microsimulation model can be used as the base population for other individual level models such as dynamic microsimulation which moves the synthetic population through space and time or agent based modelling for either enriching the model environment or to produce the acting agents 1 1 Idealised example To exemplify how the process works a simple idealised example is worked through below Table 1 shows the example survey population which will be used as the sample population in this example id sex age 1 m 20 2 f 30 3 m 30 4 f 20 5 m 30 Table 1 Sample population There are two constraint tables age and sex The constraint table for age is shown in Table 2 and sex is shown in Table 3 Both tables contain a zone field identifying the geographic zone for the aggregate counts for each row Each row represents one zone Each field following the zone field contains information relating to the attributes within that constraint For Table 2 the first field contains the counts of people aged up to 20 for each zone the second field contains the counts of people aged between 20 and 30 For Table 3 the counts of males m a
9. is greater the change is rejected As the algorithm proceeds down the optimisation path and the annealing schedule reduces the likelihood of accepting a change for the worse reduces Once a change has either been accepted or rejected the maximum number of improvement attempts and the maximum number of improvements made are checked to see if either limit for these progress counters has been reached If the limits have not been reached the algorithm suggests another random change and continues processing from this stage looping back to the change suggestion stage until one of the limits is reached or a perfect fit for the fitness calculation is achieved When one of the limits is reached the temperature is reduced by multiplying it by the annealing factor this multiplier is set by the user and controls how quickly the temperature threshold will reduce and therefore how quickly the algorithm will move towards a solution Once the temperature has been adjusted the count of annealing stages the large outer loop is checked to see if the maximum number of stages has been reached If there are more stages left the counters for maximum number of attempts and maximum number of improvements are reset and the algorithm loops back to suggest another random change If there are no more stages left the algorithm has reached the end 28 Start Generate random parameters and test fitness Suggest a random change to parameters
10. Source tab at the top right hand side of the main screen highlighted in Figure 5 below Processes Winda Add data source Figure 5 Data sources tab Right click anywhere in the white area to see the context menu to register a data source displayed in Figure 5 As the Add data source option is entered a second menu with the data source types becomes visible currently the only option is to read and write flat files such as csv Selecting the Flat File option will open a window in the main action area of the screen for importing data from a flat file format Figure 6 The application links to the data source storing information about the location and format of the file rather than copying the data into a local area 11 ij Flexible Modelling Framework version 1 0 Beta Application Microsimulation N Import Data From Flat File Select a directory lt lt No Data Location Selected gt gt is Sohimech Figure 6 Import data window To link to a data source you can either type directly into the text area denoted by lt lt No Data Location Selected gt gt or click on the button with on it to launch a directory chooser dialog window If you choose to type the location of the data source directory into the text area and the path or directory is incorrect then the statement could not connect to data source lt lt data source name entered gt gt will appear in the reporting area at the
11. aggregate counts for each type of person in each zone hereon referred to as constraints e To evaluate how well the population has been created verification tables are required This is an optional but recommended post population creation stage Tables not used as constraints but used available to evaluate the population will be referred to as evaluation tables 6 2 1 Structure of the sample population The format for the sample population is to have one record for each unit person household type Each record should have a unique identifier the P_ID field in the example below A field for each of the desired constraints and evaluation attributes should also be present Gender Age Ethnicity Age_Gender etc in the example below Figure 29 25 li Table Population Se ange Year ee Pon Ase PA ee Ethnicity Age_Gender Age_Ethnicity Gender _Ethnicity 10001 Males ic o white _ British 0_to_15 _Male_People 0_to_4 White_British Male_ White_British a Males f z K EF White_British lo _to_15_Male_People ox to_4_White_British Male_ White_British Males 2 0 to_15 O_to_4 O_to 4 White_British 0_to_15_Male_People 0_to_4_White_British Male_ White_British Males 3 O_to_15 0 _to_4 O_to_4 White_British 0_to_15 Male People 0_to_4 White_British Male_White_British Males 4 0 to_15 0_to_4 0 _to_4 White_British O0_to_15 Male People 0_to_4_White_Briti
12. and test fitness Reset Apply If change attempts and Metropolis algorithm accepted increment improvement to accept or reject improvement counters change counter Maximum improvement attempts tried or maximum improvements made Increment attempts counter Reduce annealing factor More annealing stages Output final results End Figure 32 The Simulated Annealing algorithm There are three ways the algorithm can reach the end 1 when all iterations have been exhausted and an optimal solution has not been found 2 if when the fitness is tested after a change the fitness statistic represents a perfect fit O for the Total Absolute Error statistic the algorithm will break straight to the end no matter what stage the different counters and limits are at 3 if the algorithm either makes a complete circuit of the outer annealing loop without taking an improving step this is difficult to do if a backwards step is taken or 10 outer loop iterations are completed without the fitness statistic changing the algorithm will exit The latter conditions have been 29 included to allow the algorithm to exit early when it has reached a point where no further progress is made and a perfect solution has not been reached 6 4 What all the bits of the screen mean Figure 33 below shows the Microsimulation configuration screen Each area of the screen is highlighted and explained
13. configuration lt lt model name gt gt Table for zone totals x 10 IV Randomise Run Figure 41 Saving the link noting that any fields from the constraint table not included in relationships with the sample population will not be included in any model run Once a link has been saved right clicking on it in the population table area will present the same context menu we used to start and add the link above this time the Set ID field and Add a link options are disabled and the Remove link and Edit link options are enabled Figure 42 Selecting the Edit link option will reload the relationships and table information into the link table area allowing adjustments to the relationships to be made Selecting the Remove link option will delete the link from the configuration altogether One link can be defined for each of the fields in the sample population table except for the field reserved as the unique identifier but links do not have to be defined on all fields in the table Only one constraint table can be linked to a field If the field names in the constraint table and the attribute values in the sample population table are not the same suggestions will not be made by the application and the user needs to manually create the relationships as shown in Figure 43 below for the Gender constraint ij Microsimulation Population Table Population Optimisation settings set the maximum number of s
14. in more detail below All of the controls on the screen have help text associated with them which appear if the mouse is hovered over the item Flexible Modelling Framework version 1 0 Beta Application Microsimulation ij Microsimulation k E Population Table Optimisation settings set the maximum number of steps in the optimisation algorithm _ 100 set the maximum number of improvement attempts before continuing lt _4 100 set the number of improvements to the optimisation before continuing lt lt lt lt _ _ 10 set the annealing reduction Factor recomended 0 9 0 9 Output to lt lt output data source gt gt B lt output table group name gt gt __ Save configuration configuration EEE name gt gt Figure 33 Microsimulation configuration screen 1 Population table area is where the sample population table for creating the synthetic population is configured 2 Link table area is where each of the constraint tables is configured and the direct relationship between each attribute in the sample population table and the fields in the constraint or link tables are defined 3 Drop down box for identifying the constraint table to be used to calculate the over population in each zone If all constraints sum to the same value for each zone the ideal situation this is not critical to set and can be left as the first defined constraint In some cases the popula
15. registering a table A dialog box asking if you wish to permanently delete Age appears Figure 21 Click yes to remove the registration of the table this will NOT remove the underlying data file Click no to cancel the operation and retain the table registration fjDrop Table From Msim Figure 22 Un register table confimation dialog 4 3 Loading data and opening tables To load data into a table without displaying the data right click on the table and select the Load data option The table will have a green tick through it to show that data has been cached in memory successfully shown below in Figure 22 20 Processes windows Data Sources P Msim Se a Properties z File Name Tworki FostDoci Popula Le Data Type Flat File Tables ee Age need Age by Ethnicity Tiri iB Ethnicity S Gender Gender_by_Age Population Sex _by Ethnicity Figure 23 Data cached in sample tables Right clicking on the table when it has data loaded shows the context menu below Figure 23 giving an option to Clear cache Selecting this option will remove the data from memory and return the table to a normal state P Msim Be Properties Led File Name Ci Work PostDoc Popula Data Type Flat File Hg 496 a S table EE Ger EE GE HE Fop Drop table Sexe by Ethnicity Figure 24 Table context menu All context menus produced from right clicking on a table have an option to Ope
16. value can be entered If you try to run a model with a non integer value as the seed a message The seed is not a valid Long number please correct and try again Will be displayed in the reporting area and the model run will be aborted The settings for the optimisation parameters default to 100 100 10 0 9 It is recommended that you explore these with regard to your own model configuration as each simulation problem demands a different amount of work from the algorithm and these settings are relatively arbitrary When the run button is pressed the microsimulation screen is closed and the optimisation process begins The Processes tab should now display an overall progress bar called Optimising highlighted in red in Figure 59 below This progress bar shows the increments through the whole model run when 100 is reached the model run is finished The progress bars highlighted in yellow show the progress through the optimisation for each zone in turn The application will use as many processing cores as are available therefore if the computer has two cores available two zones will be optimised in parallel if four cores are available four zones will be optimised in parallel Therefore the more cores available on a computer the faster the model will complete To cancel the overall model run click the Cancel Optimising button The zones being optimised can either be allowed to complete or the cancel buttons underneath
17. 1 42 43 43 44 44 45 45 46 46 47 48 1 Introduction to static spatial microsimulation Microsimulation as an analytical approach dating back to the work of Orcutt 1957 and Orcutt et al 1961 The modelling approach constructs a dataset of individual units persons households or firms etc over a large area To accomplish this microsimulation uses a sample population to copy or clone individuals to match constraints specified using known aggregate data Therefore the distribution of constraint attributes in the synthesised population should match those observed in the constraint data Static spatial microsimulation takes this approach and effectively adds a strict spatial constraint to the synthesis process Populations are created within smaller geographical areas specified by the user This spatially explicit approach has had a number of diverse applications including the analysis of disease prevalence in health studies Brown amp Harding 2002 Smith Pearce amp Harland 2011 Tomintz Clarke amp Rigby 2008 transportation analysis Beckman Baggerly amp McKay 1996 McFadden Cosslett Duguay amp Jung 1977 and studies associated with demand estimation such as Williamson and Clarke 1996 water demand estimations Static spatial microsimulation uses survey data to sample from and aggregate data Census data for example to constrain the sampling process The resulting synthetic population contains the
18. 48 Figure 49 Figure 50 Figure 51 Figure 52 Figure 53 Figure 54 Link table area populated with field and values from constraint and sample population Saving the link Link context menu in sample population table area Manually configuring a link All constraints configured Selecting most trusted constraint Setting output location Saving a configuration Configuration name conflicts Saved configuration Loading a configuration Reloaded configuration Locating missing data when reloading a configuration Adjusting the location of the sample population table Altering the fields in the sample population table 25 25 25 26 28 29 30 31 31 32 32 33 34 34 35 36 36 37 37 38 38 38 39 39 40 40 Figure 55 Figure 56 Figure 57 Figure 58 Figure 59 Figure 60 Figure 61 Figure 62 Figure 63 Figure 64 Figure 65 Figure 66 Figure 67 Figure 68 Deleteing a model configuration Delete configuration confirmation dialog Model configuration deleted Selecting the random settings Model processing progress bars Output table generated Structure of the output statistics table Structure of the output population table Evaluation window Evaluation menu below the microsimulation options Completed evaluation window Create tables process running Example summary tables created Summary fit tables for the example model configuration 41 41 4
19. 6 4 Whatall the bits of the screen mean 6 5 Setting up a model configuration 6 5 1 Creating links between the sample population and constraint tables 6 5 2 Selecting the constraint to calculate total population values from 6 5 3 Adding output information 6 6 Saving loading and deleting a model configuration 6 6 1 Save 6 6 2 Load 6 6 3 Loading a configuration where data sources or tables have changed 6 6 4 Delete 6 7 Running a model configuration 6 7 1 Final setup and run 6 7 2 Output and interpretation 6 7 3 Evaluating the model 7 References o oo o ac 10 10 10 11 11 14 20 22 22 23 23 23 23 23 24 25 25 25 25 26 26 27 30 32 32 37 37 38 38 39 40 42 42 42 44 45 50 8 Suggested Further Reading 50 Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Figure 22 Figure 23 Figure 24 Figure 25 Figure 26 Figure 27 Figures Splash screen for Flexible Modelling Framework Main screen of the Flexible Modelling Framework Application menu File chooser dialog Data sources tab Import data window Open file dialog Import data window with file path specified Data source in the data sources tab Expanding the data source Data source properties Data source context menu Tables context menu
20. 8 Summary fit tables for the example model configuration 49 7 References Beckman R J Baggerly K A amp McKay M D 1996 Creating synthetic baseline populations Transportation Research 30 415 429 Brown L amp Harding A 2002 Social modelling and public policy Application of microsimulation modelling in Australia Journal of Artificial Societies and Social Simulation 5 4 6 http jasss soc surrey ac uk 5 4 6 html Harland K Heppenstall A Smith D and Birkin M 2012 Creating Realistic Synthetic Populations at Varying Spatial Scales A Comparative Critique of Population Synthesis Techniques Journal of Artificial Societies and Social Simulation 15 1 1 Available From http jasss soc surrey ac uk 15 1 1 html Kirkpatrick S Gelatt C D and Vecchi M P 1983 Optimization by Simulated Annealing Science 220 671 680 McFadden D Cosslett S Duguay G amp Jung W 1977 Demographic Data for Policy Analysis Urban Travel Demand Forecasting Project Final Report Series Vol VIII University of California Berkeley and Irvine Institute of Transportation Studies Metropolis N Rosenbluth A W Rosenbluth M Teller A H amp Teller E 1953 Equation of State Calculations by Fast Computing Machines Journal of Chemical Physics 21 1087 1092 Orcutt G H 1957 A new type of socio economic system The Review of Economics and Statistics 39 116 123 Orcutt G H Green
21. a window with file path specified Once a directory has been chosen and the connect button clicked the Import Data From Flat File will reset More data sources can be added at this point or the import window can be closed using the cross hairs in its upper right corner highlighted in Figure 8 The data source s added should appear in the Data Source tab to the left as shown in Figure 9 below Processes Windows Data Sources P Msim Figure 9 Data source in the data sources tab The black cylinder shape denotes a correct connection to the underlying data source has been established Double clicking on the data source will expand it showing the Properties and Tables that belong to this data source Currently this data source does not have any tables therefore there is a large dot displayed next to the Tables whereas the Properties has a folder and a symbol next to it Figure 10 13 Processes Windows Data Sources HES Properties Fue 5 Tables Figure 10 Expanding the data source The properties for the data source can be expanded by clicking the symbol Figure 11 The fully qualified file name is displayed and the data type although currently only Flat File is available Processes Windows Data Sources a Properties bn File Name C Work PostDoc Populatic i Data Type Flat File so Tables Figure 11 Data source properties Highlighting a d
22. ains the Application and Microsimulation menus in the screen shot above 2 The large dark grey action area below the menu ribbon which is where any action windows appear 3 The light grey tabbed pain to the right containing three tabs a Processes this is where the progress of any operations in progress can be observed b Windows a list of all open windows in the main action window are displayed here c Data Sources contains a tree of all registered data sources tables and fields 4 The white strip at the bottom is the reporting window and where any non critical information notices will be displayed 3 2 Application menu On the menu ribbon the first option is Application This menu can be dropped down by clicking on it or by holding down the Alt key and clicking A at the same time The menu options can be seen in WgFlesible Modelling Framework version 1 0 Beta Application Microsimulation Set Data Directory i Quit Processes Windows Data Sources Ctrl g Figure 3 below Figure 3 Application menu 3 2 1 Shutting down To quit the application select the bottom option Quit or click Q while holding down the Ctrl key note that the application menu does not have to be displayed for this shortcut combination to cause an exit to occur Alternatively exiting the application can be achieved by clicking on the cross hair box in the top right hand corner of t
23. aming convention of lt lt output prefix gt gt _ lt lt table name gt gt _ lt lt statistic abbreviation gt gt _gof For the age constraint and Total Absolute Error statistic in the example here this would be Eval UG _Test_Age TAE gof The structure of the table is identical to that of the original table with the first field containing the zone identifier followed by the field headings used in the link to the sample population The values for each field are the values for the fit statistic chosen for the individual cell in the table 48 These three levels of fit tables provide detailed information about where the model has performed well and poorly The list of tables created in the evaluation process can be seen in Figure 68 ae 18 x Processes Windows Data Sources P Msim EE Properties H E Tables iid T Age S Age_by Ethnicity S Ethnicity Gender Gender by Age Population hs Sex _by Ethnicity UG output HO Properties Ou Tables a SUG Test population UG _Test_stats Eval UG Test Gender S Eyal UG Test Age SS Eyal UG Test Ethnicity S Eval UG Test Gender by Age S Eval UG Test Age by Ethnicity S Eval UG Test Sex by Ethnicity Eval UG Test Gender_T4 amp E_gof S Eval UG Test Age TAE_gof S Eyal UG Test Ethnicity T amp E_gof S Eval UG Test Gender by Age TAE S Eval UG Test Age by Ethnicity TA S Eval UG Test Sex by Ethnicity Ta S Eval UG Test_zones_gof Eval UG Test_gof Figure 6
24. ata source and right clicking again brings up a context menu Figure 12 this time a few more options are available The Register file option is considered in section 3 2 below The clicking the Remove data source option will remove all record of the data source from the application This does not affect the underlying directory and files on the computer these will remain in place this action simply removes the link to the files from the application Processes Windows Data Sources BAe Register file Remove data source DociPopula Add data source b Figure 12 Data source context menu 4 2 Registering and unregistering files Within the data source we need to register the data files This is purely the process of telling the application what format the files are in so that the data can be loaded correctly The Register file option can be accessed by right clicking on the data source as shown above or by right clicking on the Tables node Figure 13 below 14 Processes Windows Data Sources eP Msim BS Properties inp File Name C Work PostDociPopula Data Type Flat File Register file Clear all table caches Figure 13 Tables context menu Once the Register file option has been selected a window called Register Flat Files lt lt data source name gt gt will be displayed in the action area Figure 14 ij rlexible Modelling Framework version 1 0 Beta
25. berger M Korbel J and Rivlin A 1961 Microanalysis of Socioeconomic Systems A Simulation Study New York Harper and Row Smith D M Pearce J R amp Harland K 2011 Can a deterministic spatial microsimulation model provide reliable small area estimates of health behaviours An example of smoking prevalence in New Zealand Health amp Place 17 618 624 doi 10 1016 j healthplace 2011 01 001 Tomintz M N Clarke G P amp Rigby J 2008 The geography of smoking in Leeds estimating individual smoking rates and the implications for the location of stop smoking services Area 40 341 353 doi 10 1111 j 1475 4762 2008 00837 x Voas D amp Williamson P 2001 Evaluating goodness of fit measures for synthetic microdata Geographical andEnvironmental Modelling 5 177 200 doi 10 1080 13615930120086078 Williamson P amp Clarke G P 1996 Estimating small area demands for water with the use of microsimulation In Microsimulation for urban and regional policy analysis Clarke GP Ed London Pion 8 Suggested Further Reading Microsimulation TANTON R Vidyattama Y Nepal B amp McNamara J in press Small area estimation using a reweighting algorithm Journal of the Royal Statistical Society Series A 50 WILLIAMSON P Birkin M amp Rees P H 1998 The estimation of population microdata by using data from small area statistics and samples of anonymised records Environment and Pla
26. bottom of the application The safest and most flexible way to register a data source is to click on the button A dialog like the one below in Figure 7 will appear The dialog will open at the folder set through the Set Data Directory option in the Application menu below this is C Work PostDoc Population Synthesis on a windows machine If the data directory set does not exist or has not been entered the dialog will open at the default location for the local machine 9 Fuzzy TAE calibration My Recent 3 Population Distribution Documents E Desktop My Documents Wr Ta My Computer T Folder name C Work PostDoc Population Synthesis Open My Network Places Files of type jal Files Cancel Figure 7 Open file dialog 12 The dialog can be navigated in the normal way with new folders being added and moving up or down levels until the desired folder has been located Once this is highlighted click open and the full path and name of the folder should be displayed in the Import Data From Flat File windows text area If the correct path is shown click connect otherwise launch the file chooser dialog again and attempt to navigate to the correct folder ij Flexible Modelling Framework version 1 0 Beta Application Micrasimulation ig import Data From Flat File Select a directory i WiorkPostDoc Population Synthesis Data Mining OukputiMsim Connect Figure 8 Import dat
27. change The fields in the table are now examined to see if they match those stored in the configuration If they do no further action is taken If the fields do not match the previously defined fields the field search dialog is displayed To change the field locate the new field location from the field list by expanding the cross at the side of the table Left click on the correct field and drag it into the grey box above the buttons it name will appear as shown in Figure 54 below and click change Once all changes have been made the configuration will be loaded If not all links can be loaded a warning will be displayed in the reporting area so that the user can take action As shown below the warning contains information on the field from the sample population table and constraint table name forming the link that has been removed to assist in any updates that may need to be made Reforming the link is simply a case of adding the link as described in the section 5 5 1 above WARNING Not all existing value to field pairs could be found E TE age ry Elric hanged Mh Age changed E w E by Age changed H Genees rhegen HETT Sax by Dthnioty dhanged E y Popidstinn changed Ea PID Field ge corneal be found in g Gender table Population_changed g Sree ear O jum Please spinel ther correct nid Age Band TET m oP Age Senden gi Age Ethncty l ae Dkr _Flbie iy e Lis_output_changed Figure 54 Altering the fields
28. constraint 6 5 3 Adding output information The next stage of configuration is to identify where the model outputs should be saved First identify the data source where the tables should be saved in the data sources tab click on it and while holding down the left mouse button drag it into the lt lt output data source gt gt area as shown in Figure 46 below Next to the data source is a free text box to enter a prefix that all model outputs will use to identify them in this case UG Test 37 set the mxdmum number of stegs in the optimisation algcethm set the macdmum number of improvement attempts before continuing w set the number of inpeovements to the optimisation befoen continuing j 19 set the annoaling reduction factor recomended 0 9 FH os S a fe Save configuration 10 T Randomese lt cmodel name gt Figure 46 Setting output location 6 6 Saving loading and deleting a model configuration 6 6 1 Save To save a configuration enter a name in the highlighted area of the screen in Figure 47 in this case User Guide Ex and click the save button This action will save all the configuration settings entered so far under the name specified i Microsimulation Population Table Population Fields ingle_Year_OfF_Age ge_Band ge_PA wn Linked Field Optimisation settings set the maximum number of steps in the optimisation algorithm 100 set the maximum
29. d 6 3 How the algorithm works The algorithm is an iterative optimisation algorithm simplified from simulated annealing Kirkpatrick et al 1983 which in turn is an implementation of the Metropolis Algorithm Metropolis et a 1953 A flow diagram representation is shown in Figure 32 below The optimisation algorithm is executed for each zone individually It starts by randomly sampling a population of the correct number of people households from the sample population table The fitness of the population is then calculated measured as the Total Absolute Error equation 1 Voas and Williamson 2001 between the count of individuals in each category in the synthetic population and the expected count from the constraint tables summed across all constraint tables 27 i j A change to the synthetic population is suggested a randomly identified individual from the synthetic population is replaced by a randomly selected person from the sample population The fitness of the synthetic population is again calculated If the change improves the population fitness the change is automatically accepted If the change degrades the fitness the exponential of the fitness difference a negative number divided by the temperature the distance travelled down the optimisation path is compared to a randomly generated number between 0 and 1 If the randomly generated number is less than the exponential of the change calculation the change is accepted if it
30. d Linked Field inked Field Linked Field Linked Field Linked Field Gender Ethnicty Linked ied Figure 44 All constraints configured 6 5 2 Selecting the constraint to calculate total population values from Once all of the links are added the constraint which is most likely to hold the actual number of population units for each zone is selected in the drop down box Table for zone totals at the bottom of the screen In this example we are using the Gender constraint because it has the smallest number of categories male and female and should therefore be subject to the least amount of adjustment for disclosure control Figure 45 ii Microsimulation i Population Table Population Optimisation settings Pop id field set the maximum number of steps in the optimisation algorithm lLinked field 100 set the maximum number of improvement attempts before continuing Linked field _ 100 beac Field Age_Gender Linked Field set the number of improvements to the optimisation before continuing Link Table 10 set the annealing reduction Factor recomended 0 9 0 9 Output to lt lt output data source gt gt lt lt output table group name gt gt Save configuration lt lt model name gt gt Table for zone totals age X 10 JV Randomise Age Ethnicity Gender Gender_by_Age Age_by_Ethnicity Sex_by_Ethnicity Figure 45 Selecting most trusted
31. data source and enable the user to revert the name or location back to that expected by the application If the directory is returned to the original name and location and contains all of the tables previously registered then the application will automatically validate the data source and tables the next time it restarts currently restarting the application is the only way to restore the link to the data in these circumstances without re registering the data source and all tables 22 If the data source directory or path is changed externally while the application is open tables will still be able to be opened but they will be empty and when closed or saved the current defined path in the application will be created with an empty file representing the table closed saved 4 4 2 Change in file name If a file name csv file containing data changes whether it is before or after the application is opened the table will still appear in the data source It will behave as described above if opened it will contain no data and when closed or saved a new file will be created using the table name in the data source directory 4 4 3 Change in file structure If the file structure of the file you are trying to load does not match that registered in the application the statement The file you are trying to load lt lt table field name gt gt is not registered correctly fields lt lt count of fields gt gt registration lt lt count of registe
32. each of the progress bars can be clicked to cancel each zone optimisation process that has been left running When the model completes the approximate time taken is displayed in the reporting area 43 Processing Time is Weeks 0 days 0 hours 0 minutes 17 seconds 55 millie seconds 156 Optimising GODAFA0040 2 Cancel Optimising OOD 4FA0040 Optimising OOD4FA0045 Done Gancel Optimising DUD4Fa0n4s Optimising OODAFAOON46 1 Cancel Optimising OOD4F 40046 Figure 59 Model processing progress bars Processing Time is 18 minutes 6 7 2 Output and interpretation Results are saved into the data source specified in the output area of the microsimulation screen Two tables will be created with the prefix entered on the microsimulation screen called lt lt prefix gt gt _population and lt lt prefix gt gt _stats In the example used here the tables are UG_Test_population and UG_Test_stats Figure 60 Processes Windows Data Sources P Msim EE Properties H E Tables e Ua output BH Properties eRe Steep lables H E UG Test _population fl UG Test stats Figure 60 Output table generated The population output table contains a list of all of the cloned person records along with the zone identifier the person record has been cloned into The structure of the population tables is just two fields one contains the zone code from the constraint tables and the other contai
33. er of improvements made before progressing the thresholds for the inner loop Increment attempts counter Reduce annealing factor Optimisation settings set the maximum number of steps in the optimisation algorithm Th e la st S id er f 10 set the maximum number of improvement attempts before continuing annealing reduction More annealing stages factor the amount J the temperature is set the number of improvements to the optimisation before continuing reduced by a fte r each inner loop is Output final results com plete set the annealing reduction Factor recomended 0 9 Tae Figure 34 Relationship of Simulated Annealing algorithm and configuration controls to move towards accepting improving steps only and settle towards a solution 5 Output area is the registered data source where the output tables will be stored and the prefix that will be used to identify the tables created by this model run 6 Save configuration provides the opportunity to name the configuration and save it for later use or edit 7 Randomise settings are where the user can specify the model to be run in random mode or provide a seed to be used Specifying a seed to be used ensures the results are reproducible 31 so long as the same input data and random seed are used with the same version of the algorithm 8 The run button will start the microsimulation model 6 5 Setting up a model configuration
34. fferent valid constraint category combinations such that all distributions within the data can be considered and simulated by the 26 microsimulation algorithm Some combinations are obviously not valid we would not expect to find children in an employment category for example Each constraint should relate to one field in the sample population table and each category in the constraint that has a count of greater than O for any zone should be represented in the corresponding field in the sample population table Constraints should not relate to more than one field in the sample population and sample population fields should only refer to one constraint or evaluation table A relationship between two categories in the sample population table and the corresponding constraint table is highlighted in Figure 31 It is worth noting that both the sample population and constraint tables contain many more records than displayed in the figure ij Table Population PID Gender Single_Year Age_Band Age_PA Age Ethnicity Age_Gender Age_Ethnicity Gender _Ethnicity White _ British 0 _to_15 Male People 10_to_4 _White_British Male_White_British a White _British 0_to_15_Male_People 0_to_4_white_British Male_white_British white _British 0_to_15_Male_People 0_to_4_white_British Male_White_British White_British O_to_15 Male People 0_to_4_White_British Male_White_British
35. figuration lt lt model name gt gt Table for zone totals 10 JV Randomise Figure 37 Pop id field displayed in action To add a link we locate the field we wish to constrain in the population table as we did above to set the ID field left click to highlight it and then right click to show the context menu again This time we select Add a link from the context menu The example in Figure 38 shows a link being added to ij Microsimulation Population Table Population Optimisation settings set the maximum number of steps in the optimisation algorithm 100 set the maximum number of improvement attempts before continuing Linked Field w Pop id field Drag the population table from the data source tree and use the context menu to create and ec Link Table 10 e a a a ee set the annealing reduction factor recomended 0 9 Ee is Output to lt lt output data source gt gt lt lt output table group name gt gt Save configuration lt lt model name gt gt Table for zone totals s 10 IV Randomise Run Figure 38 Link added to population table the Age field Notice that the text Linked field is now displayed in the Action column and the Save configuration and Run buttons are disabled additionally the context menu accessed by right clicking in the population area is no longer available This is because we have only half finished c
36. g Framework version 1 0 Beta Application Microsimulation ig Register Flat Files Msim Age csy Age_by_Ethnicity csyv Ethnicity csv Gender csy Gender_by_Age csy Population cs Sex_by_Ethnicity csv ETT sey ses wen ee a a a a E oara zal Oo o a E aa al E a E 0DAFA00OS 10 16 ODAF A006 40 8 14 ODAF A000 37 19 21 ODAFA000S 12 13 31 OODAFA0009 44 19 26 4 Figure 16 Register flat file headers delimiter and text qualifiers 16 Notice that the headers are now named as contained in the first row of data A selection of common file delimiters and characters used to enclose text fields can be chosen in the area highlighted to the top right in Figure 16 The default for these two options is a comma for the delimiter and double quotes for text encapsulation this is the format used to save all flat files generated by the application The field types will be suggested by the application from the sample of top 10 rows If a text value is found in these rows the type suggested for the field is text otherwise it is suggested as a numeric of double precision Notice in Figure 15 the sample data is all left aligned in the columns indicating text This is because the first row does indeed contain text for the field headings In Figure 16 when the first row is used as a header only the first field has text detected and is therefore left aligned all other fields are suggested to be numeric and are aligned
37. he application main screen highlighted in Figure 3 above 3 2 2 Setting the data directory The user has the option to set the root data directory Clicking on Set Data Directory will launch a directory chooser dialog box shown below in Figure 4 Simply navigate to the directory on your computer that you would like to use as the root source for your data selection and click open Please note that this is a convenience option to eliminate long repetitive file searches and can be overridden when actually selecting your data This option simply means that when data is being selected the selection dialog will open at your preferred data directory This option can also be changed at any time without any adverse effects 10 Look in H My Computer r Ti E El See Local Disk C3 e C all Users My Recent 3 Shared Documents ocuments i Audio CD D E ae KINGSTON H i Desktop hy Documents Wr My Computer N Folder name Open My Network Places Files of type fal Files Cancel Figure 4 File chooser dialog 4 Data Sources Before we can use the application to do any modelling activities we need to link to the data required to undertake the activities In the application there is the concept of a data source This is the location of data to be used in a modelling process or a location where data is to be saved 4 1 Registering and unregistering a data source To register a data source click on the Data
38. in the sample link between Gender and table Gender_changed has been removed link between Age_Gender and table Gender_by_Age has been removed When the configuration has loaded a reminder that it has changed and that the changes have not yet been saved will be displayed 41 Changes have been made to the configuration User Guide Ex save the configuration to ensure changes are available in the future 6 6 4 Delete To delete a model configuration go to the Delete microsimulation model option on the main Microsimulation menu Hover over the option until it expands and then click on the model name you wish to delete Figure 55 ij rlexible Modelling Framework version 1 0 Beta Application Microsimulation New microsimulation model Load microsimulation model Delete microsimulatioan model User Guide Ex User Guide Ex to delete Mew validation setup Load validation b Delete validation Figure 55 Deleteing a model configuration When you have selected the model configuration to delete a confirmation dialog like the one in i Microsimulation 2 Do you want ta delete User Guide Ex to delete Figure 56 Delete configuration confirmation dialog Figure 56 below will be shown To cancel the delete operation click No to carry on and remove the configuration click Yes Once deleted the model configuration will no longer be available to load or delete as sh
39. ion tables ii Microsimulation b Population Table Validation settings Population to evaluate lt lt result table gt gt Zone ID Field lt lt not set gt gt Person ID Field lt lt not set gt gt Standardized Root Mean Square Error 7 Absolute Entropy Difference I R squared Entropy standard entropy measure lige I Total Absolute Error Standard Absolute Error Output to lt lt output data source gt gt lt lt output table group name gt gt Save configuration Figure 63 Evaluation window There are three main differences to the main microsimulation screen 1 The area at the top of the Validation settings pane highlighted in red above The top box in this area is where you enter the table name for the output population created by the model in the example UG _Test_population the second box is for the zone identification field in the example ZonelD and the third box is for the unique person identification field in the example P_ID 2 The area highlighted in yellow is for the selection of fit statistics to be evaluated against the most useful of these is the Total Absolute Error measure which is near the bottom of the list Simply locate the statistics you wish to use and check the boxes next to them 3 The third difference is the three buttons at the bottom of the screen instead of one This is so that different fit statistics can be added without having to run all of the evaluati
40. lease make sure it is installed before trying to start the application The software can be downloaded from https github com MassAtLeeds software releases 3 Starting the software Starting the software can either be accomplished by double clicking on the exec jar file if jar files are set to execute on the machine you are using If double clicking causes no action it is likely that jar files are not associated with java to create the association on a windows machine see If configuring jar files to execute with java cannot be achieved start a command window navigate to the folder where the exec jar file has been copied and type in java jar exec jar and press return As the application starts the splash screen shown in Figure 1 will be displayed while all of the code modules are loaded and checked MASS Group University of Leeds loading module Microsimulation jar Figure 1 Splash screen for Flexible Modelling Framework Once the application has successfully loaded the main screen will be visible Figure 2 ir lenble Modeling ranvework wersion LU Weta olf x Applicaon Microsimulation Flexible Hodelling Framework version 1 U Peta 15 an open source application developed at the University of Leeds Figure 2 Main screen of the Flexible Modelling Framework 3 1 Navigating the screen The screen has four main areas 1 The menu ribbon across the top of the screen cont
41. n User Guide Ex Table for zone totals Gender v 10 IV Randomise Run Figure 49 Saved configuration If the configuration name already exists a dialog like the one shown in Figure 49 will be displayed Clicking No will cancel the save operation allowing the name to be changed on the main microsimulation screen before saving is attempted again Clicking Yes will overwrite the saved version of this configuration with the one that is currently displayed on screen iia Microsimulation lt r A model by this name already exists overwrite Figure 48 Configuration name conflicts 6 6 2 Load To retrieve a previously saved model configuration simply navigate to the Load microsimulation model in the main Microsimulation menu and hover over the option until it expands and then select the model configuration name you wish to load Figure 50 ij Flexible Modelling Framework version 1 0 Beta Application Microsimulation New microsimulation model Load microsimulation model User Guide Ex Delete microsimulation model User Guide Ex to delete New validation setup Load validation Delete validation Figure 50 Loading a configuration Clicking on the named model will load up the microsimulation screen with the options specified in the configuration as shown in Figure 51 below 39 Flexible Modelling Framework version 1 0 Beta Application Microsimulation ij Microsimulation User Guide Ex Pop
42. n Table Population Validation settings Pop id field Population to evaluate UG_Test_population Linked Field Zone ID Field ZoneID Person ID Field P_ID lLinked field Linked Field J Total Absolute Error Age_Gender Linked field x 7 Standard Absolute Error Link Table Percentage Error pills te Total error Cell Percentage Error Output to UG_output Eval_UG_ Test UG Example Load Data Figure 65 Completed evaluation window Loading the data enables the Create Tables button Once the Create Tables button is clicked a progress bar will appear in the Processes tab showing how far through the summary creation the application is Figure 66 le x Processes Windows Data Sources Creating summary For OODAGEOOS6 Cancel Creating summary For OODAGE Figure 66 Create tables process running 47 This will take some time especially if there are many zones or a large population The summary tables created for the example used here are shown in Figure 67 below x i Processes Windows Data Sources P Msim E E Properties Se a Tables E iii Age E S Age by Ethnicity EH T Ethnicity E TT gender E S Sender _by_Age E Population H Sex by Ethnicity a US output E E Properties BM UG Test population gg ZonelD bo of PID UG Test stats Gf Eval UG Test_Gender E Eval UG Test Age Fb Eval UG Test_Ethnicity E
43. n it is removed and replaced This is behaviour by design files not registered and visible in the application are not deleted If a file is registered and visible in the application it is assumed that the user would like to overwrite the information 6 7 3 Evaluating the model It is possible to evaluate the model fit The options for doing this are present in the lower half of the Microsimulation menu Figure 63 The Load validation and Delete Validation options work in the Same way as the Load microsimulation model and Delete microsimulation model options explained above Once configurations have been created they are accessible through the menu areas 45 ig rlexible Modelling Framework version 1 0 Beta Application Microsimulation Mew microsimulation model Load microsimulation model Delete microsimulation model Mew validation setup Load validation Delete validation Figure 64 Evaluation menu below the microsimulation options To create a new validation configuration click on the New validation setup shown in Figure 63 and the screen show in Figure 64 will open The screen is very similar to the main microsimulation screen and entering the population and creating links is done in exactly the same way as described in section 5 5 You are not limited to the same links as used in the microsimulation if other information is available to be checked these data can be included here as evaluat
44. n table Selecting this option will load data into memory and open a window containing the data in the action area of the screen Figure 24 Opening a table can also be achieved by left clicking on the table and dragging it into the action area while holding down the left mouse button 21 Flexible Modelling Framework version 1 0 Beta Application Microsimulation ij Table Age _by_ Ethnicity gooaragon sel E O O E e S S a S S a S E S opaFaooos av gooaraooor i S S qopaFaooos E S S omaras e o o n a te S S owaraooit E T S S A e E E S S e S E S S waa ao o e aas aoo o a 4 F Rows 2439 Figure 25 Table window opened in the action area of the screen If you select to Clear cache of a table that is open in the action area it is automatically closed If you have several tables with data loaded you can right click on the table folder and select Clear all table caches This will remove the data from memory for all tables in the table folder 4 4 What happens if file names or directories change between loads 4 4 1 Change in data source directory If the path or directory name of a data source changes between loads then the verification process for the data source will fail and it will be displayed as shown below in Figure 26 F Msim Figure 26 Invalid data source The data source information can still be expanded to allow the user to find the location that is being searched to find the
45. nd females f are shown for each zone Both tables contain information for three zones a b and c The total population for each zone in this simple example is very small zone a 20 zone b 10 and zone c 12 Zone 20 30 a 7 13 b 4 6 C 9 3 Table 2 Age constraint table Zone m f a 5 15 b 5 5 c 4 8 Table 3 Sex constraint table The constraint tables can be matched against the fields in the sample population in this example the fields in the constraint tables match the attributes within the sample population perfectly i e the sex field in the sample population has values of m and f which match the fields in the constraint table the same is true of the age table In reality this is rarely the case and pre processing is generally required to get the constraint and sample population information to match The static spatial microsimulation process works through a zone at a time and clones individuals from the sample population into the synthetic population while matching the sum of individuals attributes to the constrained counts For zone a the synthetic population may look like that contained in Table 4 Zone id sex age a 1 m 20 a 1 m 20 a 1 m 20 a 1 m 20 a 1 m 20 a 4 f 20 a 4 f 20 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 a 2 f 30 Table 4 Example synthetic output for zone a Counting the individuals in the example solution for zone a sho
46. nd in table Gender changed Please select the correct field Field name changed dialog change cancel Figure 52 Locating missing data when reloading a configuration Clicking cancel on either dialog will cancel the change to the location of the table or field If the table or field is in the sample population table this will result in the load of the configuration being cancelled as a model run cannot be executed without the sample population table being present If cancel is pressed during the search for an altered constraint table then the loading of the link associated with the constraint will be cancelled and the link will be removed from the configuration although the configuration will still be loaded This is also the case if the values contained in the 40 sample population table s field cannot be matched to the fields in the constraint table as previously mapped To make an update to a table locate it in the data sources area left click on it and drag it into the grey box above the buttons the name of the table will appear in the box as shown in Figure 53 Applicaton Microsamulation j jSciect correcttabe x Table Population cannot be found in data source CaVorkiPosMociPopulation Synthesis Oata MininglOutputMsin Please select a valid table pulation_ehanged Figure 53 Adjusting the location of the sample population table below When you are happy you have selected the correct table click
47. ndow in the Windows tab will also make the selected window active and move it to the front of the display if it is hidden by any other windows If more than one window is opened with the same name a number is appended to the end of the window title in the form 1 as shown in Figure 27 below for the window Microsimulation Microsimulation 1 and Microsimulation 2 The window title shown in the top banner of the window in the activity area will always match the title displayed in the Windows tab 23 3 wl 35 s G x2 E s 55 6 8 59 8 Pa 6 Ei ry z 9 x 37 49 ed eoa Flexible Modelling Framework version 1 0 Beta if an Open source application developed at the University of Leeds Figure 27 Interaction between windows tab and open windows in the action area 5 2 The processes tab When the application is working processes that are currently being executed are displayed in the Processes tab as shown in Figure 28 below The blue bar is the progress bar which increments as a process moves towards completion In the progress bar is displayed the process name and beneath each progress bar is a related cancel button When a process is live the cancel button will be enabled but once the process reaches the end the button will be disabled and the progress bar and button will then be removed from the Process tab Clicking the cancel button indicates to the application tha
48. nning A 30 785 816 MERZ J 1991 Microsimulation A survey of principles developments and applications International Journal of Forecasting 7 77 104 Simulated Annealing INGBER L 1993 Simulated Annealing Practice Versus Theory Mathematical and Computer Modelling 18 29 57 51
49. ns the population identifier from the sample population table Figure 61 Using an external database the original sample population table can be joined back to the output population table using the unique population identifier in this case P_ID with the location for each synthesised person being identified by the zone code in this case ZonelD 44 ii table UG_Test_population zoned PID podAFAooa foooe1 OODARAGOOL floo0zs3 aopAFACaG foo014s ooDAFAOOOL ro00z6 oODAFAOOO1 1000147 JOODARAOON1 JOODAF A001 JOODAFAOON1 QODAFAOOOI ToooIs JOODAF AGO JOODAF A000 Rows 715402 Figure 62 Structure of the output population table The stats output table contains the fitness level recorded at each major iteration of the optimisation algorithm for each zone It therefore has three fields the first is the zone identification field ZonelD the second is the count of the iteration for which the fitness value was output Order and the third is the fitness Total Absolute Error value itself Fit Figure 62 ij Table UG_Test_stats E Bl x Lie oe 00DAFB0027 42 6 JOODAFEOO7 i ee a e 5 6 4 i il Rios Bese Figure 61 Structure of the output statistics table Note if an underlying file for the output table being created already exists but is not registered in the application the values are appended into the file If the file exists and is registered in the applicatio
50. ntains all of the fields from the constraint table The Value column contains all of the attribute values found in the associated field in the sample population table The application will attempt to match the field name to attribute where the two match It is worth making sure that this has been done for all of the relationships between fields and values that you want to make The top row in the constraint table is empty in the Value side of the display because no match was found for this field this is expected as this is the zone identification field We need to tell the model that this field is the zone identification field We do this by left clicking in the empty Value cell to reveal a drop down of all the possible attributes values from the sample population field with the option for Zone ID at the top Figure 40 select this option i Microsimulation Population Table Population Optimisation settings set the maximum number of steps in the optimisation algorithm 1 100 set the maximum number of improvement attempts before continuing 100 set the number of improvements to the optimisation before continuing m 10 set the annealing reduction factor recomended 0 9 a Pop id field Output to lt lt output data source gt gt lt lt output table group name gt gt L J Save configuration lt lt model name gt gt Table for zone totals
51. number of improvement attempts before continuing thnicity ge_Gender ge_Ethnicity Gender_Ethnicity Link Table Linked Field Linked Field Linked Field Linked Field lt 100 gt set the number of improvements to the optimisation before continuing a m 10 set the annealing reduction factor recomended 0 9 PH os Output to Save configuration User Guide Ex Table for zone totals Gender V Randomise Figure 47 Saving a configuration The name for the configuration should now appear in the expanded areas next to the option to Load microsimulation model in the main Microsimulation menu as shown below in Figure 48 38 fg Flexible Modelling Framework version 1 0 Beta Application Microsimulation New microsimulation model Load microsimulation model gt User Guide Ex Delete microsimulation model gt Ox Population Optimisation settings New validation setup set the maximum number of steps in the optimisation algorithm Single_Yea 4ge_Band Load validation _ _ 100 Age_PA Delete validation Age TTET set the maximum number of improvement attempts before continuing Ethnicity Age_Gender N 100 Age_Ethnicity Linked Field Gender_Ethnicity Linked Field iy set the number of improvements to the optimisation before continuing Link Table 10 E a Perrine ee rene emer F w Save configuratio
52. o_4 gl 5_bo_ al B_to_9 a 10_to_14 SNE age_15 ol 16_bo_17 ool 18_bo19 a 20_to_24 ool 25_bo_29 ol 30_to_44 ool 45_bo 59 al 60_bo_64 ol 65_to_74 ol F5_bo_84 ool 5_bo_89 Figure 19 Expanded table showing field types The registered file is no longer available in the Register Flat Files window as shown below in Figure 20 the Age csv file is no longer visible The last files registration details remain set in the window and the Register button will be disabled until another file is selected to be registered 18 fig Flexible Modelling Framework version 1 0 Beta Application Microsimulation ig Register Flat Files Msim ge_by_Ethnicity csy Sex_by_Ethnicity csv ae ee EA A a a a E A E E A E E oara ooo o aoo o a o soar aon D i _ a E aol a Register Figure 20 Files already registered no longer show in file selection area To unregister a table right click on the table in the Data Sources area to show a context menu demonstrated below in Figure 21 Select the bottom option Drop table 19 Processes windows Data Sources eP Msim E Properties ji File Mame C Work PostDoc Popula 4 Data Type Flat File H Tables Open table Load data Drop table ae 16 to 17 ai lt P 18 to 19 sa gt 20 to 24 ol 25 bo 29 a g 30_to 44 oo 45_bo 59 DNE 60 to 64 AT g 65_to 74 l AT lt P 75 to 84 ane W a5 to 69 Figure 21 Un
53. on stages again after they have been run once The three stages have been separated purely for convenience Re running the Create Tables option is time consuming and is not required if the model has not been re run or the number of links in the evaluation has not been 46 changed since a previous creation of the summary tables If either of these conditions have changed links or model run then all three stages will be executed e Load Data ensures that all of the data used to create the tables is loaded correctly e Create Tables only becomes enabled when data has been loaded This option creates a summary set of tables in the same structure as the constraint evaluation tables using aggregate counts from joining the sample population table to the output synthetic population e Run calculates the fit statistics and creates the evaluation tables requested If additional fit statistics are required after an initial evaluation has been run and the summary tables have already been created the process can be shortened Simply select all of the required fit statistics click Load Data and once data loading is complete click the Run button missing out the time consuming Create Tables stage The completed Validation screen for the example used here can be seen in Figure 65 below Flexible Modelling Framework version 1 0 Beta Application Microsimulation i Microsimulation UG_Example Populatio
54. on the row with the field name in to highlight it and then right click to show the context menu as shown in Figure 36 below Select Set as ID field and the text Pop id field should appear in the Action column as shown in Figure 37 i Microsimulation l Population Table Population Optimisation settings EEF set the maximum number of steps in the optimisation algorithm Set as ID field 100 Add a link Remove link set the maximum number of improvement attempts before continuing Edit link _ 100 Age_Gender set the number of improvements to the optimisation before continuing Link Table 10 ek ye oead a eoi 0 9 Output to lt lt output data source gt gt lt lt output table group name gt gt Save configuration lt lt model name gt gt Table for zone totals X 10 JV Randomise Figure 36 Selecting the unique identifier ii Microsimulation Population Table Population Optimisation settings set the maximum number of steps in the optimisation algorithm _ _ 100 set the maximum number of improvement attempts before continuing _ 100 Pap id field set the number of improvements to the optimisation before continuing Link Table m 10 e l set the annealing reduction Factor recomended 0 9 0 9 Output to lt lt output datasource gt gt lt lt output table group name gt gt Save con
55. own below 6 7 Running a model configuration 6 7 1 Final setup and run ij Flexible Modelling Framework version 1 0 Beta Application Wicrosimulation New micrasimulation model Load microsimulation model P Delete microsimulation model User Guide Ex Mew validation setup Load validation Delete validation Figure 57 Modet2onfiguration deleted To complete the final setup select whether you would like the model to run using in a random mode or whether you would like to supply a seed to ensure the results can be recreated using the same data seed and algorithm The randomise options are in the bottom right of the microsimulation screen Figure 58 g Microsimulation User Guide Ex E Population Table Population Optimisation settings set the maximum number of steps in the optimisation algorithm Pop id field Linked field j1 100 ji pj set the maximum number of improvement attempts before continuing Linked Field 100 Linked Fel Linked field hs m set the number of improvements to the optimisation before continuing Link Table 10 E E E T AEA 0 9 Output to UG_output UG_Test Save configuration User Guide Ex Table for zone totals Gender Figure 58 Selecting the random settings If the randomise option is selected the model will run in random mode Unchecking the Randomise option will enable the text box next to it Any valid integer
56. reating the link We have identified the field in the sample population that we want to link to a constraint table but we have not yet identified the constraint table or the attribute category to constraint field relationships To do this locate the constraint table to be associated with this population field in the data sources area and left click on it hold down the mouse button and drag the table into the Link Table area The Fields and Value columns should become populated as shown in Figure 39 below 33 ij Microsimulation Population Table Population Optimisation settings Fields tion set the maximum number of steps in the optimisation algorithm 100 set the maximum number of improvement attempts before continuing 1 1 100 set the number of improvements to the optimisation before continuing c 10 set the annealing reduction Factor recomended 0 9 Se 66 Output to lt lt output data source gt gt lt lt output table group name gt gt 7 Save contiguration lt lt model name gt gt Table For zone totals 10 V Randomise Run Figure 40 Link table area populated with field and values from constraint and sample population The label at the top of the area now shows the table name in this example the label reads Link Table Age because the age constraint is called unsurprisingly Age The Fields column co
57. red fields gt gt will appear in the reporting area of the screen To change the registration of the table you will need to right click on the table and unregister it and then re register it with the application or replace the underlying data file with one of the correct field format If text is found in a field defined as a numeric field the following statement will be displayed in the reporting area of the screen The input file format for table Age is incorrect Possible cause text characters in fields defined as number fields The cause would need to be investigated and the data either corrected or the file unregistered and re registered with the correct field types defined 4 4 4 How Null is treated Null values in underlying data will be allocated either an empty string or a O depending on whether the field is defined as a text or numeric type respectively 5 Windows and Processes 5 1 The windows tab The middle tab on the tabbed pain at the right hand side of the screen is called Windows This tab shows a list of all open windows in the action area of the screen The window that is currently active is highlighted in the windows tab as shown in Figure 27 below Table Ethnicity is the name of the window that is on top and active in the activity area and is also highlighted in the Windows tab Clicking on a different window in the action area will change the currently active window alternatively selecting the wi
58. sh Male_White_British Males 5 O_to_15 50 15 IS to 7 white British 0_to_15_Male_People S_to_15 White British Male_White_British Males le jO_to_15 S_to_15 5S to 7 White_British 0_to_15_Male_People S_to_15_White_British Male_White_British Males 7 O_to_15 5_to_15 5 to 7 White_British O_to_15_Male_People 5_to_15_White_British _ Male_White_British Males l O_to_15 S_to_15 8_to_9 White_British O_to_15_Male_People 5 _to_15 White British Male_White_British Males 9 O_to_15 S_to_15 8_to_9 White_British O_to_15_Male_People 5 _to_15 White British Male_White_British Males 10 O_to_15 5_to_15 10_to_14 White_British 0_to_15 Male People S_to_15_White_British Male_White_British Males 11 jO_to_15 S_to_15 10_to_14 white British O_to_15_Male_People 5_to_15_White_British Male_White_British Males 12 jO_to_15 5 to_15 10_to_14 White_British O_to_15 Male People S_to_15 White British Male_White_British Males 13 0 to_15 S_to_15 10_to_14 white _British 0_to_15_Male_People S_to_15 White British Male_White_British Males 14 O_to_15 Stoel s 10_to_14 White_British O_to_15 Male People 5 _to_15 White British Male_ White_British 100016 Males 15 0 to_15 S_to_15 age_15 White_British O0_to_15 Male People S_to_15_White_British Male_White_British 100017 Males 16 16_to_19 16_to_29 16_to_17 White_British 16_to_19_Male_People 16_to_29_white_British Male_Wwhite_British Individual unique identifier Figure 29 Example sample population 6 2 2 Struct
59. t the particular process is no longer required it is worth noting that the process may take some time seconds through to a few minutes depending on the complexity of the work being undertaken to register the cancelation exit and remove the process from the tab If an error occurs during the execution of a process the text in the process bar will report the error and a message will normally be displayed in the reporting area and a detailed stack trace recorded in the log file in the root folder where the application has been deployed 24 Verifying Data Sources Done Gancel verhying Data Sources Checking Msim Done ance checking Msim Figure 28 Progress bars for processes in the processes tab 6 Microsimulation 6 1 What is the technique and what does it do Static microsimulation is the process of creating a detailed population from aggregate data reproducing the relative known distributions within the population as accurately as possible Several approaches can be taken and are discussed in the wider literature for one example see Harland et al 2012 The implementation discussed here uses a combinatorial optimisation approach called Simulated Annealing see further reading for suggested explanatory texts 6 2 Data format To create the synthetic population the following are required e a population to sample from hereon referred to as the sample population e constraint tables containing information about the
60. teps in the optimisation algorithm FH 100 set the maximum number of improvement attempts before continuing Linked field SENA w Add a link Pop id field Age_Gender set the number of improvements to the optimisation before continuing Remove link Link Table Edit link c 10 set the annealing reduction factor recomended 0 9 0 9 Output to lt lt output data source gt gt Save configuration lt lt model name gt gt Table for zone totals age 10 V Randomise lt lt output table group name gt gt Figure 42 Link context menu in sample population table area Add the constraint table to the link table area but no field value relationships automatically identified Identify the field containing the zone unique identifier in the constraint table Manually identify the Male field to Males attribute relationship Manually identify the Female field to Females attribute relationship Figure 43 Manually configuring a link 36 The process of linking the sample population fields to the corresponding constraint tables is repeated until all of the required constraints have been added In the example used here three univariate constraints have been added Gender Ethnicity and Age and three cross tabulated constraints have been added Age by Gender Age by Ethnicity and Gender by Age Population Table Population Fields op id Fiel
61. tion is most reliable in one table and is less reliable in other tables for example the 2001 UK Census has disclosure control applied to tables which mean that not all constraint tables will add up to the expected total Pre processing could be undertaken to resolve this situation or alternatively this drop down box can be used to select the most likely constraint to contain the correct population counts for each zone 4 Optimisation control area Each slider is used to control an aspect of the simulated annealing algorithm Figure 34 below demonstrates how the adjustment sliders relate to the flow diagram introduced earlier Increasing the values for each of the limits will mean extends the algorithms potential to cover more search area Increasing the annealing 30 reduction factor reduces the amount that the temperature decreases on each major iteration This increases the search space covered by the algorithm as it takes longer for it Start Generate random parameters and test fitness Suggest a random change to parameters and test fitness The top slider sets the number of annealing stages the outer loop in each optimisation Reset Apply If change attempts and Metropolis algorithm accepted increment improvement to accept or reject improvement counters change counter The second slider controls the number of improvements sna attempts and the third the epee he numb
62. to the right to show this To override the suggested typing of the field you can right click anywhere in the data area for a field and a context menu showing the current data type will appear and allow you to alter this to a different type if required Figure 17 J Headers in First row gODAFAQ002 E g eoo o o a y loobarA0003 sting OODAFAOOO4 Integer JOODAFAONOS Bache JOODAFAOOOE loobaranoa a7 oaran 4 E Figure 17 Data type context menu Once all settings have been correctly selected click the Register button in the bottom right of the window and the file will be registered and should appear in the Tables area of the data source Figure 18 Note the Tables icon has now changed to a folder and it has been expanded in the same way as explained for the properties option above O CUCU la x Processes windows Data Sources e Msim a Properties RN File Mame Ci w orkiFPostDociPopula Data Type Flat File Ho Tables lo Age Figure 18 Registered table appears in data sources tab 17 Clicking on the cross to the right of the table Age displays the fields in the table and how they have been defined as demonstrated below in Figure 19 Each field has an icon displaying either abc or 123 next to it demonstrating if the field type is registered as a text field or a numeric field P Msim E Properties beg File Name C WorkPostDoc Popula al O_t
63. ulation Table Population Optimisation settings Pop id field set the maximum number of steps in the optimisation algorithm Linked field 100 set the maximum number of improvement attempts before continuing Linked field J w Linked field Age_Gender Linked Field Ly set the number of improvements to the optimisation before continuing Link Table 10 re w J set the annealing reduction Factor recomended 0 9 0 9 Output to UG_output UG_Test Save configuration User Guide Ex Table For zone totals Gender v 10 JV Randomise Figure 51 Reloaded configuration Note the optimisation settings randomise selection status and seed settings are not stored in the configuration and need to be adjusted each time a model is run 6 6 3 Loading a configuration where data sources or tables have changed When a data source or table has been moved renamed or altered you will be asked to locate the new location for the table before the configuration can be loaded as shown below If the structure of the table has changed or field names are different you will be asked to locate the missing fields Figure 52 fj select correct table Table Population cannot be found in data source WyorkiPostDociPopulation SynthesisiData MiningwOutpunn sin Please select a valid table Table name changed dialog change cancel i select correct field x Field Male cannot be fou
64. ure of the constraints and evaluation tables Constraint tables should contain a zone field and counts for each category within the attribute for each zone All constraint tables should contain the same number of zones and have the same zone identifiers although the order of the zone records within the tables is unimportant The structure of a constraint table is shown in Figure 30 below gender by age The left field in the example is the zone code identification field and the remaining fields are all categories for individuals within this constraint containing the count of individuals within each category It is important that all of the fields containing counts for the constraints are registered as numeric fields Mj table Gender_by_Age 33 25 5 5 Attribute headings 40 25 5 el 45 43 7 i a 12 ol o 32 21 3 5 38 34 2 5 55 gl 1 34 8 1 59 8 26 6 29 z 8 23 24 5 9 30 21 6 5 37 31 6l 3 1 49 48 12 10 1 5 Al 6l 0 Zone identifier It is worth noting that both constraint and evaluation tables are structured in exactly the same way Figure 30 Constraint table example structure the only difference between the two tables is that evaluation tables do not necessarily need to be included as constraints during the model run 6 2 3 How the tables relate The sample population table should contain all of the di
65. ws that we have 5 males 15 females 13 20 to 30 year olds 30 and 7 individuals aged 20 or below 20 Table 4 is only one configuration of cloned individuals from the sample population that match the constraint counts There are many different configurations for a simple problem such as this that would match the constraints However it is unlikely that any real world model would be as simple as this and as the model complexity increases the number of suitable solutions decreases compounded by issues of real world data quality Furthermore to increase the model accuracy cross tabulated attribute information can be used as the constraint increasing the information captured in the model and passed to the resulting synthetic population In this example this would involve using counts of individual for all age and sex combinations m 20 f 20 m 30 and f 30 It is easy to imagine how the number of constraining attributes can escalate quickly increasing the complexity of the model It is for the researcher to establish a balance between model complexity and accuracy 2 Installing the software Installing the software is just a matter of copying all of the files in the distribution onto your local machine The software does not need administrator privileges to install you just need to have read and write access to the area of the machine where you copy the program files to The software requires the Java Runtime Environment jre6 or later to operate p

Microsimulation model user guide

Contents

Download Pdf Manuals

Related Search

Related Contents