Publication as PDF - the Max Planck Institute for Demographic
Contents
1. b Fertility 1 Life Expectancy Ay age Av age Ay age Percent one Percent two 2 Life table P ae at bith ESE child children 22 18 25 63 23 65 0 0725 0 2367 0 69 3 Age specific fertility 4 Cohort fertility statistics win Other tables 5 Unions by parity 6 Births by union status amp order File View Group View Figure 4 The Modgen User Interface of the final LifeCourse model 2 4 7 BioBrowser output of individual life courses Figure 5 below displays the life course of a simulated actor experiencing a first birth before first union formation first union formation during pregnancy second birth within the first union and a union dissolution followed by a second union We recommend the use of the BioBrowser tool for debugging whenever new events and or status variables are added to the model in order to verify proper status changes in all status variables 26 Modgen BioBrowser Biog1 bbr Ss 10 x S File Edit View Format Tools Filter Window Help l x olele all m ala 2 Filter tracking 90 100 Kil 4 gt r gt Current age group index gt Union status at first birth First birth in first unio First birth out of unio Ho first birt m gt Parity status Childless gt Union status at second birth A Second birth in first unio Ho second birt gt Union status A First Union gt 3 year fter fi2. 2 0 231365 0 239884 0 148317 2 0 110490 0 072949 Tm i 0 4i0 i1 94 5 0 165551 0 147390 0 108470 0 080378 0 033944 10245 9 0 2938546 0 0261186 0 0456905 3 991086 13 0 5807688 6121143 2 0 5987179 5 0 9019690 172 0 0150836 0 0110791 2 0 0370541 2 0 012775 2 0 0661157 all processes added at this point start at the occurrence of events e g the process of union dissolution begins at union formation this requires additional clocks recording 1 the time since first conception 2 the time since last partnership formation and 3 the time since last union disso lution All clocks look up their corresponding parameters determining the next cutting point in time As for age groups we allow user to change the time intervals 19 TAMA ADA AAA AAA AAA ALAA AAA ALAA AAA LALA LALLA MAI AAA ALLAM A Clock events file clocks mpp ADDITIONS STEP 3 Hi LUMA ALAA DAYAL ALT AAD U UNUU UUN UNUUNU ALA LMA EAA LAMA DALAL LAA ALA MAA AAT actor Person Individual int birthspell 0 Current interval since last birth TIME lastbirthspellchange TIME INFINITE Time of last interval index change event timeNextBirthSpellEvent NextBirthSpellEvent Event setting index last birth int unionspell 0 Current interval since union form TIME lastunionspellchange TIME INFINITE Time of last interval index change event timeNextUnionSpel
3. Union status at first birth PARITY STATE parity status CHILDLESS Parity status int parity 0 Parity TIME lastbirth TIME INFINITE Time of last birth event timeFirstBirth FirstBirth First birth event he PUAT ALE TAAL LAMA AMAA AAA ATA ONU UVUA UU AAA AAA TEAL SAA ATA eee stem suratela TIME Person timeFirstBirth TIME event time TIME INFINITE double randdur hazard if parity status CHILDLESS hazard First_age agespell First_unionStatus union status if hazard gt 0 randdur log RandUniform 1 hazard event time WAIT randdur return event_time void Person FirstBirth parity lastbirth time parity _status ONE_ CHILD if union_status FIRST UNION PERIOD1 union _status FIRST UNION PERIOD2 first status FIRST IN FIRST UNION else if union _status SECOND UNION first status FIRST_IN SECOND UNION else first _status FIRST NOT IN UNION 2 3 4 Adding table output We add two tables in order to report fertility The first displays the simulated first birth and second birth rates as well as overall fertility by age It makes use of the Modgen function entrances which counts the number of actors entering a specified level of a specified state variable The function value_out returns the value of a variable when leaving the cell this expression is used to calculate cumulative rates The second table reports cohort measures namely the averag
4. Max Planck Institut fiir demografische Forschung Max Planck Institute for Demographic Research Konrad Zuse Strasse 1 D 18057 Rostock GERMANY Tel 49 0 3 81 20 81 0 Fax 49 0 3 81 20 81 202 http www demogr mpg de MPIDR WORKING PAPER WP 2006 046 NOVEMBER 2006 The LifeCourse Model a competing risk cohort microsimulation model source code and basic concepts of the generic microsimulation programming language Modgen Martin Spielauer spielauer demogr mpg de This working paper has been approved for release by Vladimir M Shkolnikov shkolnikov demogr mpg de Head of the Laboratory for Demographic Data Copyright is held by the authors Working papers of the Max Planck Institute for Demographic Research receive only limited review Views or opinions expressed in working papers are attributable to the authors and do not necessarily reflect those of the Institute The LifeCourse Model a competing risk cohort microsimulation model source code and basic concepts of the generic microsimulation programming language Modgen Martin Spielauer November 2006 Abstract This paper documents the source code of LifeCourse a simple competing risk microsimulation model initially developed alongside a study on fertility decline in Bulgaria and Russia LifeCourse is programmed in the generic microsimulation language Modgen developed at Statistics Canada In the co
5. TALE MLD ALAA AIAG AMAL AAA AAD AA YADA AMALIA A A LA EEI LAM LAELIA MI AGAMA EM ALLA AIAG ed classification LIFE STATE ALIVE NOT_ALIVE he range LIFE 0 40 parameters logical canDie double ProbMort LIFE parameter group P01 MORTALITY canDie ProbMort he partition YEARS 2001 2002 2003 2004 2005 AGILG AOL AGL AGL ZOA0 AOS 2032 AV32 AGS 2035 he if hil Hil if 2007 2022 2037 Life status Alive Dead Simulated age range Swith to turn mortality before age 40 on off Age specific death probabilities a Mortality parameters Single Calendar Years 2008 2009 2010 2011 2012 2023 2024 2025 2026 2027 2038 2039 2040 RON 2028 2014 2015 2029 2030 8 Parameter groups are listed alphabetically in the Modgen application To control their order we use the name prefixes PO1_ P02_ etc The actual parameters are stored in a dat file which is loaded at runtime The syntax is similar as in the mpp file except that actual values are assigned to the parameters In our case we initialize canDie with the logical value TRUE and ProbMort by a comma separated list of death probabilities Modgen recognizes repeaters placed in parenthesis 40 01 is equivalent to 40 times 01 Note that Modgen applications automatically include a visual user interface that allows manipulating model parameters in concise tables For la
6. FIRST UNION FORMATION 0 03 Before first pregnancy 1 0 13 In first pregnancy 18 38 0 17 first year after birth Tees ayes year 2 3 after birth 0 42 0 15 ear 4 after birth 0 29 0 11 Before 1989 BASE 1 0 08 1989 1993 0 94 0 03 1994 1998 0 69 0 05 1999 0 56 0 04 The process for first union dissolution starts at first union formation The model controls for different episodes of motherhood and calendar time 30 FIRST UNION DISSOLUTION Ist year of union union duration 1 5 union duration 5 9 union duration 9 13 union duration gt 13 before first pregnancy during first pregnancy first 3 years after birth 3 6 years after first birth 6 9 years after first birth 9 years after first birth 0 01 0 02 0 02 0 02 0 01 1 0 29m OSS 0 61 0 60 0 90 Before 1989 BASE 1989 1993 1994 1998 1999 The process for second union formation starts at first union dissolution The model controls for different episodes of motherhood and calendar time SECOND UNION FORMATION lt 2 years after dissolution 2 6 years after dissolution 6 10 years after dissolution 10 15 years after dissolution 15 years after dissolution Before first pregnancy In first pregnancy Mother 0 20 0 14 0 11 0 03 0 05 1 SO Es 0 58 Before 1989 BASE 1989 1993 1994 1998 1999 Second union dissolution is modeled as the process star
7. Parameter Groups a Mortality parameters Swith to turn mortality before age 40 on off Age specific death probabilities Table Groups a Life tables 1 Life Expectancy 2 Life table Reading Parameters 100 IS Table 2 Life table Transformation Value Rows in 2001 2001 2002 Aunn LS Parameter Swit LS Parameter Age m Model Run Progress d a 2 LS Table 1 Life Expectancy Transformation Value Columns Selected Quantities 10000 Columns Output for each calendar year Selected Quantities Population at Average start of population in 10000 2004 2005 File View Group View Ready Figure 2 The user interface of Modgen applications 2 2 8 Displaying individual life courses by the BioBrowser program The Modgen Biography Browser BioBrowser is a tool for the graphical display of individual life courses This view on simulation results is especially useful for model debugging In order to use the tool the tracking feature has to be switched on in the scenario settings and the list of variables to be tracked has to be declared within a track actorname code block in an mpp file the tables mpp file being a good location track Person current_age life status Modgen tracks all changes of state variables included in the tracking statement for a sample of simulated actors In our case we
8. by parity_status The second table displays the number of women unit by first_status and second_status i e union status at first and second birth In order to use the BioBrowser tool we include a track statement for the automatic tracking of the most important variables for graphical output UU UWUR UNUU UUU UNUU ALAA DAA TAA AGA ELAS AAA ATA AAG IAM IAA TAA MAA AAT AEA UUNIN UUU IE Output tables tables mpp ADDITIONS STEP 3 Hah TTA ALGAE AEA OA TEA LA EAL OAL TEA ELLY AEA EEL OAL table Person T05 UnionsByParity life _status NOT ALIVE 5 Unions by parity unit Cases value _in nUnions unit Number unions in life decimals 4 parity status 1 table Person T06_BirthsByUnion life_status NOT_ALIVE 6 Births by union status amp order first_status second status unit he track Person current_age life status parity agespell birthspell unionspell dissspell nUnions parity status union status first _status second status he 2 4 6 Running the model After having successfully converted the mpp code into c code by Modgen and having built the Visual Studio Solution we can run the final version of our model The selection window of the resulting application now contains the full list of model parameters and all six tables defined previously 25 I5 lifecoursed C C Programs TechPaper_Step3 Base sce Scenario Edit view Window Help F D0 U8 e i we
9. from corresponding birth out of union to birth within union levels 11 More precisely she enters the first period of the first union FIRST_UNION_PERIOD1 from where she is automatically moved to FIRST_UNION_PERIOD2 after three years by the union duration clock if the union was not dissolved in the meantime 22 EULA LALA LALLA AAA AMA ALAA DA AMAA ALAA AAA AE LAA AAA Hi Union formation amp dissolution events unions mpp introduced at STEP 3 PUMA LAAN ALA ALA AAA ALAA ALAA MAA AEA LADLE A TAAL AAA actor Person Hi int nUnions 0 fi TIME lastunion TIME INFINITE fifi TIME lastdiss TIME INFINITE fii event timeUnionlFormation UnionlFormation Teh event timeUnionlDissolution UnionlDissolution Hf event timeUnion2Dissolution Union2Dissolution Hi event timeUnion2Formation Union2Formation Val he Individual Union counter Time of last union formation Time of last union dissolution First union formation event First union dissolution event Second union dissolution event Second union formation event PAM MAMA TABLE ALAA EA AA AAALAC EAL EAE EE EAA AEA TEE First union formation TIME Person timeUnion1lFormation TIME event time TIME INFINITE double randdur hazard if union_status NEVER_IN UNION hazard Forml1_age agespell1 if parity _status ONE CHILD hazard hazard Form1_spellBirth birthspell1 if hazard gt 0 randdur log RandUniform 4 hazard ev
10. the integer variable current_age The current_age variable is a characteristic of the actor Person and defined within an actor Person code block It is initialized with 0 In this block we also find the declaration of the birthday event methods introduced by the Modgen statement event Each event has two methods the first determines the timing of the event the second determines the consequences of the event For any given point in time the timeBirthdayEvent method returns the time of the next birthday To determine the event time we make use of the Modgen function WAIT which adds a specified duration the age at the next birthday minus the current age to the current point in time When defining clock events we need to consider the nature of competing risk models i e that the first occurring event censors all other events This effect is not wanted for clock events where we allow more than one clock to ring at the same point in time To assure that no clock event is censured each clock has to be able to determine whether or not it is time to ring right now and in order to avoid loops to record the last clock event Accordingly we track the time lastbirthday in the birthday clock As the actual day of birth at age O is not considered a first birthday we increase current_age only if the age is above 0 Alternatively current_age may be initialized with 1 HAA MAMA ALM AAA EAL A AAA UNUN UU UUN UV UUU UNUUNU UMUN
11. D STATE second status NO SECOND Union status at second birth event timeSecondBirth SecondBirth Second birth event i LATE LATA MMA MALIA MAA ALAA EMA AAD EMMA ALAA MMA AEA Second birth TIME Person timeSecondBirth TIME event time TIME INFINITE double randdur hazard if parity status ONE CHILD hazard Secon_age agespell Secon_spellBirth birthspel1 Secon_unionStatus union status Secon firstStatus first status if hazard gt 0 randdur log RandUniform 2 hazard event _time WAIT randdur return event_time void Person SecondBirth parity parity status TWO CHILDREN if union_status FIRST UNION PERIOD1 union_status FIRST UNION PERIOD2 second _status SECOND_IN FIRST UNION else if union_status SECOND UNION second_status SECOND_IN_ SECOND_UNION else second_status SECOND_NOT_IN_ UNION 2 4 4 Unions The main additions of the third step to our model code development are partnership behaviors first and second union formation and dissolution If a women s union_status is NEVER_IN_UNION she is at risk of entering a first union Being in a union also means being at risk of union dissolution All related events update the union status and record the time of the event The number of unions is counted by a variable nUnions which is later used for table output If a union formation occurs during pregnancy the union status at birth is ex post corrected
12. E_STATE with the possible levels ALIVE and NOT_ALIVE a state variable live_status of the type LIFE_STATE will be defined in the mortality mpp file Ranges are used to define the size of an array state variable or model parameter In our application we have defined a LIFE range with possible values from 0 to 40 Our mortality parameter is accordingly defined as double ProbMort LIFE Partitions define the breakpoints for continuous state variables and are also used for the definition of table output e g for different calendar time periods In our application we have defined a partition YEARS in order to allow the output of period tables see below One of the powerful features of Modgen is the recognition of comments placed in the programming code as labels used in the user interface of the Modgen application and in the BioBrowser tool For instance the comment Age specific death probabilities will be used as a label of the corresponding parameter list entry and parameter table See Figure 2 The parameters can be grouped for visual selection in the Group View tab of the application We defined a parameter_group PO1_MORTALITY containing our two parameters Again the comment Mortality parameters is used as a label in the application See Figure 2 LAGE MALLET EG AGA AAG AE LAL LAA AAG MAID IMAL LAE AD ALA EL A MAME LM AMAA MALL UUU UNUN Parameter definitions parameters mpp STEP 1
13. OH LS Table 5 Unions by parity Parameter Groups Ga Mortality parameters Swith to turn mortality before age 40 on off Age specific death probabilities b First birth parameters 1 0076 1 1160 Age baseline hazards first birth Relative risks of union status on first birth c Second Birth Parameters Age baseline hazards second birth Relative tisk time since last birth on 2nd birth Relative risks of union status on second birth Second birth Relative risk union status at first birth on 2nd i nm d First partnership formation parameters Second birth out of Age baseline hazards first union formation Rel risk time since last birth on 1st formation e Second partnership formation parameters Baseline time since dissolution of 2nd formation Rel risk time since last birth on 2nd formation f First partnership dissolution parameters Baseline time since formation on 1st dissulution Rel risk time since last birth on 1st dissolution g Second partnership dissolution parameters Baseline time since formation on 2st dissulution h Definition of age and duration spells Lower borders of age groups Lower borders of time in union intervals Lower borders of time since last birth intervals Lower borders of time since last dissolution Table Groups LS Table 4 Cohort fertility statistics a Life tables
14. Person NextUnionSpellEvent unionspell lastunionspellchange time if unionspell gt 2 amp amp union_status FIRST_ UNION PERIOD1 union_status FIRST_UNION_PERIOD2 ELA ELL EGU EE ELLE ELLE M E ISL L ELL I LLA LAELIA ELI LL LL EL LEVEL ELE LLEVA LLAMA ELLIE AG ILG Event that changes index of time interval since last union dissolution TIME Person timeNextDissSpellEvent TIME event time TIME INFINITE TIME spelldurat time lastdiss shige aLa for i MAX DISSOINDEX i gt 0 i if spelldurat lt dissoSchedule i amp amp spelldurat gt 0 event_time WAIT dissoSchedule i spelldurat else if spelldurat dissoSchedule i amp amp lastdissspellchange time event Cine Ciner re CUrnTevent i Cime void Person NextDissSpellEvent dissspell lastdissspellchange time 2 4 3 Second births The programming of the second birth event differs only slightly from that of first birth A woman is at risk if she has parity 1 Besides a duration baseline since last birth the model controls for age union status at first birth and the actual union status The second birth event updates the parity_status to TWO_CHILDREN and increments the variable parity Furthermore it records the union status at second birth 21 BUPA MM GALA A TAMA AA MME LA AA AAALAC AAA ALAA ALE Fertility fertility mpp ADDITIONS STEP 3 MUN A AAALAC ALAA AAA LAMA actor Person SECON
15. U NUUNUU UUU UU UNUN NUU Clock events file clocks mpp STEP 1 V VAY AMAT ADA TAD AMAL AAG AEA AAD AAA AAMAS AAA AA AAA UUN UVULA UN UUU UUU UNUUNU UM actor Person Individual int current_age 0 Current age TIME lastbirthday TIME INFINITE Time of last birthday event event timeBirthdayEvent BirthdayEvent Birthday event he TAMA AAA LAA DAA ALAM ALAA AA ALIA AA AA I AA T IV Birthday event TIME Person timeBirthdayEvent if lastbirthday time amp amp age int age amp amp age gt 0 return time else return WAIT int age l age void Person BirthdayEvent current_age lastbirthday time 2 2 4 Model parameters parameters mpp and Base lifecourse dat Our simple model template has only two parameters the logical parameter canDie a switch to turn on off mortality before the 40 birthday and age specific death probabilities The parameters are declared within a parameters code block Modgen supports the numeric types int long float and double Boolean variables logical in the Modgen terminology and a set of Modgen specific types not used in our application see the Developer s Guide p 24ff The dimensionality of the parameters is as with state variables defined by classifications ranges and partitions Classifications define all possible levels of a variable In our application we have defined a classification LIF
16. ath Event he TIME Person timeDeath TIME event_time TIME INFINITE double randdur if canDie FALSE event time else if ee a else if ProbMort int age gt 0 randdur log RandUniform 3 if randdur lt int age l age return event _time WAIT MAX LIFE age age gt MAX LIFE i event_time time log 1 ProbMort int age aL f event _time STATT randdur void Person Death life status NOT_ALIVE Finish 2 2 6 Defining table output tables mpp Modgen provides a very powerful and flexible cross tabulation facility to report model results the description of which far exceeds the scope of this paper model developers consult the Developer s Guide p 34ff Generally table definitions start with the table keyword and describe the content of table cells and table dimensions Our template model produces two simple table outputs The first contains summary values of our simulation and has no dimensions i e cells apply to the whole population over the whole simulation period We make use of the Modgen keyword unit which counts the number of actors entering the cell of a table in our case the simulation and the Modgen function duration which sums up the time actors stay in this cell in our case the total years lived by all actors in the simulation The average age at death of all actors in the simulation is then obtained b
17. ch is used by Modgen to determine an adequate rounding precision in the handling of continuous time Omitting this command may result in anomalies at model execution time PUMA AAMT UAE EAT AG ALT AAA LAA AAA MA AGAMA AAA ETD ALLA AA UMA TIAA ALI AEE LTA LUA LLAMA AMAL ALA TAL Main simulation engine lifecourse mpp introduced at STEP 1 LUMA MGA LAA AD LD EAA LLAMA DAMA EAM A UMA LIAM ALAA LEELA UMA LA AA AMAL ELA A version 1 0 Model version model type case based Model type time type double Continuous time model actor Person Individual void Start Initializes a new person void Finish Destroys the person he void Person Start time 2000 Start time of simulation void Person Finish void PreSimulation SetMaxTime 2040 Maximum value of time variable void CaseSimulation Person prPerson prPerson new Person prPerson gt Start while gpoEventQueue gt Empty if gbCancelled gbErrors gpoEventQueue gt FinishAllActors else gpoEvent Queue gt WaitUntil gpoEventQueue gt NextEvent gpoEventQueue gt Implement DeleteAllPersonActors void Simulation long 1Case for 1Case 0 1Case lt CASES amp amp gbInterrupted amp amp gbCancelled amp amp gbErrors 1Case StartCase CaseSimulation SignalCase 2 2 3 Adding a clock clock mpp The clock mpp file defines a birthday event that increases
18. e age at first and second birth and the percentage of women either staying childless or of parity 1 or 2 The table definition makes use of the Modgen value_at_transition function which in our case returns the age at parity transitions for a full list of Modgen derived state expressions see the Developer s Guide p 19ff 10 As second births are not modeled yet the tables will only report first births at this step 16 HEA TAA EAA LAD A LAD AAA UTD ELAM ALA LAMA AAA UUU UUU UNUNUN UNUNUN UNUU UNUUNU Output tables tables mpp ADDITIONS STEP 2 iy PUMA LALA ALAN ALMA ALA ALL ALA LATTA AMAA LA AAA LAMA LALLA EMAL AMAL A table Person T03 FertilityByAge 3 Age specific fertility split age AGE GROUP Output for each single age year parity duration Age specific fertility decimals 4 entrances parity status ONE CHILD duration First birth rate decimals 4 entrances parity status TWO CHILDREN duration Second birth rate decimals 4 value out parity duration Cum average parity decimals 4 he table Person T04 CohortFertility ij Cohorkerertilsityasitatlstaes value _at_ transitions parity status CHILDLESS ONE CHILD age entrances parity status ONE CHILD Av age at 1st birth decimals 2 value_at_ transitions parity _status ONE CHILD TWO CHILDREN age entrances parity status TWO CHILDREN Av age at 2nd birth decimals 2 value at _changes parity s
19. e model template After having successful run the Modgen tool for code conversion and building the C program we can test our template application The visual interface of all Modgen applications consists of two parts a selection window containing a list of all model parameters and output tables and a frame in which all corresponding tables can be displayed Before running a model we have to open a scenario by loading a scenario file The base scenario file provided Base sce contains the initial scenario settings e g the number of simulated cases and a reference to the Base lifecourse dat parameters file used The scenario settings can be changed using the Senario Settings command By saving a scenario under a new name a copy of all dat files with the new scenario name e g NewScenario lifecourse dat is created After running the scenario all output tables are updated Note that the values displayed in the output table represent only one of several possible views on results by left clicking a table a properties sheet for the table can be accessed Among other things it allows to display distributional information standard errors and the coefficient of variation of all simulated values for a in depth description of the generic Modgen user interface see the Modgen User s Guide 11 I5 lifecoursed C C Programs LifeCourseSim_Step_03_BACKSTEPO1 Base sce Scenario Edit View Window Help 6 D Suali e i we OL
20. ek Sase 1oo 199 004 1994 1000 BASE 1989 1993 1994 1993 1999 Not in union 0 06 0 07 0 05 0 03 First union lt 3 years 1 0 94 O 74 OA eE First union gt 3 years 0254 Ook Oey O 2T Second Union 0 80 0 21 0 40 023i The model for second conception uses time since first birth as baseline hazard and in addition to the interaction between calendar time and union status it controls for union status at first birth and age 29 SECOND CONCEPTION lt 0 5 years after first birth 0 07 1 76 0 5 1 years after first birth 0 15 1 76 1 2 years after first birth 0 20 TL PAD roe 2 4 years after first birth 0 23 3 1 4 6 years after first birth 0 24 0 92 6 8 years after first birth 0 15 0 78 8 10 years after first birth 0 11 0 62 10 12 years after first birth 0 07 0 48 12 17 years after first birth 0 04 0 37 17 22 years after first birth 0 03 0 20 22 years after first birth 0 00 First Birth not in Union First Birth in first union First Birth in Second Union lt 99 T T7 BASE 1989 1993 1994 1999 1999 Not in union Deer Oe Oe ae First union lt 3 years 1 0 78 OAA OP 2 nee First union gt 3 years 0 92 0 64 O 44 0 45 Second Union 3 48 1 64 1 58 1 20 The process for first union formation starts at age 15 age baseline The model controls for different episodes of motherhood and calendar time
21. ent time WAIT randdur meicusan Eyermie_ jeans 7 void Person Union1Formation nUnions union _status FIRST_UNION_PERIOD1 lastunion time unionspell 0 Ex post correction of union status at birth for union formation in pregnancy if birthspell 0 amp amp parity 1 first_status FIRST_IN_FIRST_ UNION else if birthspell 0 amp amp parity 2 second_status SECOND_IN_ FIRST UNION 23 LAGU ALUM ILE TURAL EE AAG LL GE MELA LAI LID AEE MLL A LEGA IEE LEELA EEL AL LILLE MAE ALE GLAD ALLA A EE LEGA LEGA EGG Second union formation TIME Person timeUnion2Formation TIME event time TIME INFINITE double randdur hazard if union_status AFTER_FIRST UNION hazard Form2_spellDisso dissspel1 if parity _status ONE CHILD hazard hazard Form2_spellBirth birthspell1 if hazard gt 0 randdur log RandUniform 7 hazard event time WAIT randdur return event_time void Person Union2Formation nUnions union _status SECOND_UNION lastunion time unionspell 0 Ex post correction of union status at birth for union formation in pregnancy if birthspell 0 amp amp parity 1 first_status FIRST_IN_SECOND_UNION else if birthspell 0 amp amp parity 2 second_status SECOND_IN SECOND UNION BUTE ALUE EIU ILE AGU LM T AI ELIE ELLA MDE MLE IIE LAVELLE MLE ELLE LAGE LAIMA AGE LLAMA AEM A EMERG First union dissolution TIME Person timeUni
22. er running the Modgen tool that creates the corresponding fertility CPP file the fertility CPP file has to be added as an existing item to the project i e the C files list The programming of the first birth event is straightforward we declare a state variable parity_status and initialize it with the level CHILDLESS which is changed to ONE_CHILD by birth event If a person is at risk of first birth i e childless and the hazard rate of the current age interval is greater than 0 a random duration to the next birth event is generated Note that this event will only occur if not censured by a clock event ensuring the use of an appropriate age spell index of the piecewise constant hazard rates Besides the parity status other characteristics of the Person actor are updated at the first birth event a parity variable that counts births used later for table output and the variable first_status which records the union status at first birth used later as control variable for second births As union behaviors are not modeled yet we define a variable union_status and initialize it with NEVER_IN_UNION 15 PAA MM AMAA ALAA ALM EE AMAA LAA LALLA LAA AAA ALLA AL Fertility fertility mpp STEP 2 file introduced at STEP 2 Hi PUMA MAA LAA DAN LAA AMADA LALA AA LAA AAD AAA LAA AAA LALA D ALLA LMA AMADA actor Person Individual UNION STATE union status NEVER_IN UNION Union status FIRST STATE first status NO FIRST
23. hedule dissoSchedule he 18 LUMA MAMA DAM AAA EAA LAE Parameter values file Baseline 1l LUMA AMADA AMAA AAA parameters float unionSchedule UNIONINDEX float birthSchedule BIRTHINDEX GO O75 1 25 1 75 2 75 he float dissoSchedule DISSOINDEX double Secon_age AGEINDEX O 175290 1 76470 1 29 0LS 0 20478 3 75 diy he double Secon_unionStatus UNION STA OQ ZOZAIA i Os SALPAA W 29222 ba double Secon_spel1Birth BIRTHINDEX 0 0 067913 0 145071 0 200951 0 042124 0 025746 0 000006 hi double Secon _firstStatus FIRST STA double Forml_age AGEINDEX OF 070308987 OTS 40667 0 Tey L977 0 045454 0 040038 hi double Forml1_spel1Birth BIRTHINDEX 18 383257 A 1 VOSIAL 2 O42 hi double Form2_spel1lDisso DISSOINDEX 1995702 On 353028 O d 9SiZt9 he double Form2_spel1lBirth BIRTHINDEX Dissl1_spel1lBirth BIRTHINDEX double AQaLESs he double 0 0096017 he double he 4 0 5274971 2 0 Diss1_spellUnion UNIONINDEX 2 0 0199994 0 0213 Diss2_spel1lUnion UNIONINDEX 2 4 2 Further clocks Except for first partnership formation EULA VM A THAT A MA ADAM AAA LIE AAA ifecourse dat ADDITIONS STEP 3 dik HAGA ALAA AA DLA LEA AAD AL LAA 4 On aby Bp S Sp B fe N45 Got BoIS 9 75 W075 12 75 17 5 22 75 Or Ar Ge 10 U5 js 00000 0 91534 0 78281 0 61662 0 47991 0 37239 TE 46 3 481644 0 2922246 j
24. hnical Paper 13 National Centre for Social and Economic Modeling University of Canberra http www natsem canberra edu au publications papers tps tp13 tp13 pdf Spielauer Martin Dora Kostova and Elena Koytcheva 2006 First and second births in first and second unions a decomposition of the fertility decline in Bulgaria and Russia by means of microsimulation Forthcoming check http www demogr mpg de 28 Appendix The simulated model estimated from Bulgarian GGS data The illustrative Modgen application developed in this paper is based on simple piecewise constant event history models for first and second conception and first and second union formation and dissolution estimated from 2004 Bulgarian Generations and Gender Survey data for women born after 1950 The model was developed to study the individual contribution of the different processes to the observed drop in fertility in Bulgaria and Russia Spielauer et al 2006 The Base lifecourse dat file provided contains the parameters for the period before the Bulgarian economic and political transition in 1989 bold numbers in the tables below For this analysis we exclude non Bulgarian ethnicities and censor at age 40 and at second birth The parameters for first conception consist of an age baseline and an interaction variable of calendar time and union status FIRST CONCEPTION 0 29 0 76 0 85 0 82 0 67 0 51 0 49 0 26 0 26 Dil
25. irthday as is the case with TIME INFINITE a Modgen constant for the very distant future In order to obtain random durations from probabilities we assume constant mortality hazards within each period i e between birthdays with the exception of a death probability of 1 which leads to death immediately at the start of the age year For probabilities less than 1 a random duration can be obtained from a uniform distributed random number by the transformation randdur log RandUniform 1 log 1 ProbMort int age The Modgen function RandUniform returns a uniform distributed random number between 0 1 The function takes an integer argument used to assign a different independent random number Note that comments placed in dat files are lost when saving scenarios within the Modgen application Modgen applications automatically write back the labels defined in mpp files as comments in the dat files stream to each random number function in the code When omitted Modgen automatically writes back a unique index into the mpp file before translation into C code HA AAT MANA ALA AANA EMD LAA IAD LMA AAA AL ILA EL IAAL LAL MIA AA LAME MAASAI MAMA TAL IIIA AMAL A GY Mortality mortality mpp introduced at STEP 1 EAGT EG MAE AL AAD ATT AG AAA LAME IAAL AML ATI AGI LALLA LLAMA LALA AMAL AMAA ALAM ALA AAI ELAM AMAA YD actor Person Individual LIFE STATE life status ALIVE Life Status event timeDeath Death De
26. lEvent NextUnionSpellEvent Event setting index last union form int dissspell 0 Current interval since dissolution TIME lastdissspellchange TIME INFINITE Time of last interval index change event timeNextDissSpellEvent NextDissSpellEvent Event setting index last dissolution he WALLA TAM LA ALLA DLL LOL LLL LE A TOL LAE Event changing index of time interval since last birth TIME Person timeNextBirthSpellEvent TIME event _time TIME spelldurat ation aif for i MAX BIRTHINDEX i gt 0 i if spelldurat lt birthSchedule i amp amp spelldurat gt 0 event_time WAIT birthSchedule i spelldurat else if spelldurat birthSchedule i amp amp lastbirthspellchange time event_time time TIME INFINITE time lastbirth return event_time void Person NextBirthSpellEvent birthspell lastbirthspellchange time 20 PUMA LAA AAA A LAA LLAMA AAA A LAMA AAMT Event that changes index of time interval since last union formation TIME Person timeNextUnionSpellEvent TIME event _time TIME INFINITE TIME spelldurat time lastunion int ay for i MAX UNIONINDEX i gt 0 i if spelldurat lt unionSchedule i amp amp spelldurat gt 0 event_time WAIT unionSchedule i spelldurat else if spelldurat unionSchedule i amp amp lastunionspellchange time event_time time return event_time void
27. n status The index UNION_STATE is a classification defining possible levels of union status never in union first union lt gt 3 years after first union in second union and after second union Other classifications added are PARITY_STATE and FIRST_STATE The levels of PARITY_STATE denote childlessness one child and two children the levels of FIRST_STATE represent union statuses at first birth first birth in first union in second union out of union The variables of the three classification types introduced are defined and initialized in fertility mpp the parity_status is updated at birth event the first_status at first birth 13 TAMM UMA LALLA AAD AA LAA MA AAA LLM AAA AAA LALA AAA V A I V Parameter definitions parameters mpp ADDITIONS STEP 2 Hi EUMUNDI ALAA LALLA AAA classification PARITY STATE Parity status CHILDLESS Childless ONE CHILD One child TWO_CHILDREN Two children he classification UNION STATE Union status NEVER_IN_UNION Never in union FIRST UNION PERIOD1 First union lt 3 years FIRST UNION PERIOD2 RIESE Wiolakoxa ss years AFTER_FIRST_ UNION After first union SECOND UNION Second union AFTER_SECOND UNION After second union he classification FIRST STATE Union status at first birth NO_FIRST 1 NOw feiescite DITCH FIRST _IN FIRST UNION jf Pirs DIr CHIMIEI SE UNION FIRST_IN_SECOND_UNION First birth in second union FIRST NOT IN UNION First birth out of u
28. nal ranges of time indices and parameters to define the lower borders of the intervals for the three corresponding durations union duration duration since last birth and duration since last union dissolution We add a classification SECOND_STATE which is used to record the union status at second birth 17 HALE EE GALA AAA AAA AAA AE AA IA LAA LAA AAA AAA Parameter definitions parameters mpp ADDITIONS STEP 3 Hit WALA AEA EAE ALLEL TEAL AAPL AE LALA AEA LEAL EAT EA AEA AE AEA LEAT classification SECOND STATE Union status at second birth NO_ SECOND No second birth SECOND IN FIRST UNION Second birth in first union SECOND IN SECOND UNION Second birth in second union SECOND NOT _IN UNION Second birth out of union he range UNIONINDEX 0 5 Range of union duration index range BIRTHINDEX 0 13 Range of duration spells since last birth range DISSOINDEX 0 4 Range of duration spells since union dissolution parameters float unionSchedule UNIONINDEX Lower borders of time in union intervals float birthSchedule BIRTHINDEX Lower borders of time since last birth intervals float dissoSchedule DISSOINDEX Lower borders of time since last dissolution double Secon age AGEINDEX Age baseline hazards second birth double Secon _unionStatus UNION STATE Relative risks of union status on second birth double Secon _spellBirth BIRTHINDEX Relative risk time since last bi
29. nion he range AGEINDEX 0 10 Range of age groups indexes parameters float ageSchedule AGEINDEX Lower borders of age groups double First_age AGEINDEX Age baseline hazards first birth double First _unionStatus UNION STATE Relative risks of union status on first birth he partition AGE GROUP Age ty n Sy ip Be Go Vo Bp Dp UO Ul Wa Us U4 15 16 WV US US 20 Bil 22 28 4 2S AS Bl BA 29 30 Sil B25 B39 S45 3235p SG 37p SB 39 FO ii parameter group P02_FIRST_BIRTH b First birth parameters First_age First _unionStatus He MUNUN UUU UU UU LMA UUN ALAA AAA TAA AMAT MYA MAA MDMAA MAA AA AAA EO LAA IA AAA LAA AEM Parameter values file Baseline lifecourse dat ADDITIONS STEP 2 Wi ULE TYEE GALI V ATLA ALMA MAYA AAA OHA ELA LMA MA IOLA ELIA ILA LMA LMA ELL MMA AAA AAA TUM UNUNU UNUUNU UU UUU parameters float ageSchedule AGEINDEX Oy Us WS 20 225 5 25 A2AVcby SO 22 5 85 27 5 double First_age AGEINDEX 0 2669 O 7591 0 8458 O 8167 6727 O 5105 0 4062 0 2562 0 2597 0 1522 double First_unionStatus UNION STATE 0 0648 1 0000 0 2523 0 0648 0 8048 0 0648 he 2 3 2 Adding a second clock The second clock is used to set the correct index agespell of the age baseline hazard In difference to the birthday clock which clicks every completed year of life the second clock needs to look up the parameter ageSchedule to determine the next c
30. nly model and does not include any interactions between the simulated actors In this paper we use Modgen Version 7 31 0 0 developer s edition Modgen requires Microsoft Visual Studio NET 2003 which has to be configured for Modgen applications The applications are programmed in the Modgen language the Modgen code is then converted into the C code by a Modgen tool integrated into the Visual Studio environment before the application is compiled and linked as a C application Modgen is freely available and well documented by a Developer s Guide This contribution does not aim to replace existing documentation which we frequently reference but should be see as a complement providing a code template for continuous time microsimulation in addition to the discrete time model Simpex used to introduce Modgen concepts in the Developers Guide We do not intend to introduce the full wealth of the Modgen language as Modgen supports a much broader range of modeling approaches than covered by the LifeCourse Model 2 The illustrative LifeCourse Model 2 1 Overview In the following we describe the Modgen code of LifeCourse an illustrative cohort microsimulation model The model uses a continuous time framework It simulates a female cohort from birth to their 40 birthday and was initially developed for the study of fertility change in Bulgaria and Russia Spielauer et al 2006 Except for mortality all behavio
31. nment for the programming of microsimulation models The functionality covers the handling of model parameters table definition routines for the presentation of model output For a list of links to Statistics Canada online documents concerning the Modgen language and the microsimulation models LifePaths and POHEM please consult the References a generic user interface for the microsimulation application and a broad variety of routines simplifying microsimulation programming Modgen supports both continuous and discrete time models Continuous time is usually associated with statistical models of durations to an event following a competing risk approach Beginning at a fixed starting point a random process generates the durations to all considered events with the event occurring next to the starting point being executed and all others censured The whole procedure is repeated at the new starting point until the event death of the simulated actor occurs Continuous models are technically very convenient as they allow to add new processes without changing the models of the existing processes as long as the statistical requirements for competing risk models are met see Galler 1997 for a description of associated problems The LifeCourse model developed in this contribution is a continuous time cohort model i e all actors are born at the same point in time and no new actors are created at birth events LifeCourse is a female o
32. ntain a user interface to view and change all model parameters The parameter files belong to scenarios The filenames of the parameter files consist of the scenario name plus the parameter file name in parenthesis e g Base lifecourse dat The Modgen tool produces a log file Modgen log which contains error messages In order to provide fast access to the dat and the log files within the development environment we added them to the LifeCourse solution so that they can be opened by clicking on them in the Solution Explorer window ACTORS CPP ACTORS H app ico model h model RC PARSE INF TABINIT CPP TABINT H If not starting from a template these files have to be added to the Visual Studio Project after being created by ModGen The model developer has freedom to decide how to organize the Modgen code in different files For our purpose we have chosen the following organization The main simulation file this file has the name of the application i e lifecourse mpp such a file is obligatory and contains the definitions of the model type e g case based continuous time and the code of the simulation engine which is mostly not model specific within the defined type of models One file for each group of events behaviors distinguished in the model The template contains only two types of events mortality and the corresponding file mortality mpp and clock events clock mpp Clock events are events who
33. ntext of this contribution the model is introduced step by step as template for other microsimulation applications and as training tool for demographic microsimulation using Modgen 1 Introduction The paper documents the source code of a simple cohort microsimulation model LifeCourse initially developed for the study of fertility change in Russia and Bulgaria Spielauer et al 2006 The model is programmed in Modgen a generic microsimulation programming language developed at Statistics Canada Modgen is known mainly in connection with the large and complex LifePaths and Pohem models Population Health Model two Canadian Modgen applications developed at Statistics Canada The aim of this paper is to introduce Modgen to developers of demographic microsimulation models based on event history analysis The motivation behind the paper is twofold The first reason is the strength of the Modgen language itself which unfolds for the development of simple models too and makes Modgen an interesting language for learning and teaching demographic microsimulation The second motivation is our belief that the strength of Modgen has been left relatively unexploited by developers of demographic models and that learning barriers to the underlying microsimulation technology and approach exist which we aim to reduce with this contribution Technically the Modgen language is a superset of the C programming language providing an efficient enviro
34. of microsimulation the full field of applications is much broader ranging from theoretical studies on how good statistical methods and models replicate original data to the main practical application of microsimulation detailed socio demographic and economic projections 27 Acknowledgements am thankful to Susann Backer for careful language editing of this contribution References The Modgen programming language as well as the BioBrowser Tool were developed at Statistics Canada and can be downloaded at http www statcan ca english spsd Modgen htm Also all documentation i e the Modgen Users Guide the Modgen Developer s Guide and the BioBrowser User s Guide can be downloaded from this site e Statistics Canada Modgen Version 7 0 31 User s Guide http www statcan ca english spsd ModgenUser_EN pdf e Statistics Canada Modgen Version 7 0 31 Developer s Guide http www statcan ca english spsd ModgenDev_EN pdf e Statistics Canada BioBrowser The ModGen Biography Browser Version 3 1 User s Guide http www statcan ca english spsd BioBrowser_E pdf The Modgen applications LifePaths and POHEM developed at Statistics Canada are documented at http www statcan ca english spsd e Statistics Canada LifePaths Overview V1 1 http www statcan ca english spsd LifePathsOverview_E pdf Galler Hans Peter 1997 Discrete Time and Continuous Time Approaches to Dynamic Microsimulation Reconsidered Tec
35. ol tables mpp Error Lookup Sy C files 2 ACTORS CPP T Areaan s a nev person ACTORS H OLE COM Object Viewer le cleaning up E app ico 2 Spy 4 clocks CPP fe lifecourse CPP E model h External Tools amp MODEL RC of simulation f mortality CPP Customize fti parameters CPP PARSE INF a TABINIT CPP E TABINIT H f tables cPP Dat Files Base lifecourse dat Log File e Modgen log Mapping symbols to modules file containing Done error messages 4 Task List E Output EA Find Results 1 En Search Results TBA Solution Explorer 29 class view Ln74 Coll chi ims 4 Figure 1 Modgen integrated into the C NET programming environment 2 2 1 Organization of files The Modgen code is organized into one or usually various files These have the file extension mpp As can be seen in the Solution Explorer window Figure 1 our code template consists of five mpp files We grouped them in the MPP files folder When invoking the Modgen tool which can be accessed from the Tools menu after proper installation of Modgen these files are translated into the C code the Modgen tool creates one cpp source code file for each mpp file plus various additional C files We placed the files in the C files folder The model parameters are organized into one or more dat files Modgen applications automatically co
36. omen continue to have the status after second union The code development is organized in three steps We start from a code template which only includes a clock for birthday events and mortality The resulting simple model is used to introduce basic Modgen concepts common to all models handling of model parameters actors events table output and the graphical representation of individual life courses using the BioBrowser tool In the following steps we add the behavioral modules for fertility and union formation and dissolution and additional table output thereby aiming to introduce most Modgen programming aspects relevant for the development of this type of microsimulation models 2 2 Starting from a simple model template Step 1 In this section we develop and describe a template for continuous time case based cohort simulation models We recommend always starting one s own model development from an existing template as it contains the necessary code common to all models of a given type and it implicitly contains all project settings necessary to successfully build Modgen applications within the Visual Studio NET environment The screenshot displayed in Figure 1 was taken after opening the Visual Studio Solution lifecourse sIn contained in the Step_1 directory 4 The reduction of the model to first and second births and partnerships is a simplification which is justified in the Bulgarian context where the first two birth
37. on1lDissolution TIME event _time TIME INFINITE double randdur hazard if union_status FIRST UNION PERIOD1 union_status FIRST_UNION_PERIOD2 at risk hazard Diss1_spellUnion unionspell1 if parity gt 0 hazard hazard Diss1_ spellBirth birthspell if hazard gt 0 randdur log RandUniform 5 hazard event time WAIT randdur return event_time void Person Union1Dissolution lastdiss time union_status AFTER_FIRST_UNION LTV TL EU ALLL Y LEO T ELITE CLEA LEAL ECLA AAE MAE MATE ALE GLE T EA MAEM T UEAA GE EG Second union dissolution TIME Person timeUnion2Dissolution TIME event_time TIME INFINITE double randdur hazard if union_status SECOND UNION hazard Diss2_spellUnion unionspell1 if hazard gt 0 randdur log RandUniform 6 hazard event time WAIT randdur return event_time void Person Union2Dissolution union_status AFTER_SECOND_UNION 24 2 4 5 Tables In the final step we add two tables to the application one for the number of unions by parity the second for the number of first and second births by union status at birth Both tables evaluate the final states at the end of the actor s life This is done by including a filter statement life_status NOT_ALIVE in the table definition The average number of unions in life is obtained by the Modgen function value_in nUnions divided by units it is calculated separately
38. ral models are based on simple piecewise constant hazard regression models The models and Modgen can be downloaded free of charge at http www statcan ca english spsd Modgen htm The installation procedure is described in the Modgen Developer s Guide p 54ff 141ff Modgen has to be installed in order to run a Modgen application for users of existing models Statistics Canada provides a user edition of Modgen which can be downloaded at http www statcan ca english spsd Modgen htm parameters estimated from the Bulgarian 2004 Generations and Gender Survey data are presented in the Appendix We distinguish between the following processes Mortality parameterized by age specific death probabilities converted to piecewise constant mortality risks We allow mortality to be switched off until the 40 birthday First and second birth separate models with processes starting at age 15 and at first birth respectively The models control for age and partnership status for second births we also control for union status at first birth As we only study the process to second births there are no higher order births First and second union separate models for first and second union formation and dissolution The processes start at age 15 and at the last union formation dissolution event respectively They control for age and time spent in motherhood Union s following the second union status are not modeled i e w
39. rge and multi dimensional tables the actual model parameterization can be simplified and is less error prone when using the visual interface at runtime in our case we initialize mortality with a 1 death probability for each of the first 40 years and then leave it to the user to paste in more realistic values Comments in dat files are comments only and not used as labels i e the user can change values but not labels PANAMA TAG AA EAD AMAL AAA TAA ALAA ALAA AL EAE TAA EA TAA Parameter values file Baseline lifecourse dat STEP 1 iif PAM MAA TAG AMAA TEA ALE AAA AMAA TAA TAA AAA parameters logical canDie FALSE Swith to turn mortality before age 40 on off double ProbMort LIFE 40 01 1 2 2 5 Mortality mortality mpp This file defines the mortality event that ends the life of the simulated agent We declare a state variable of type LIFE_STATE which is initialized with ALIVE at birth and set to NOT_ALIVE by the death event The death event also calls the function finish the latter which deletes the actor The timeDeath method returns a death date If canDie is set to FALSE the date of death is always the 40 birthday otherwise the method determines by the Monte Carlo simulation if and when an actor dies before the next birthday event If the actor survives the method returns TIME_INFINITE Note that the event of birthday automatically censors the process if the time of the event is after the next b
40. rst unio First union lt 3 year After second won Hever in unio Second union gt Current interval since union form 2000 2010 2020 2030 2040 Time Database C C Programs T echPaper_Step3 Base trk mdb Filter Person X 2026 26 Ready Figure 5 BioBrowser output final model 2 5 Using the model The LifeCourse model presented in this contribution was initially developed for the study of fertility decline in Bulgaria and Russia Spielauer et al 2006 and served as a template of sister models currently developed for the study of the effect of union dissolution on fertility in France In the presented context of demographic analysis microsimulation mainly serves as a means of synthesizing different event history models in order to assess the effect of individual parameters not only on the risks of single events but also on the whole system under study For instance from the Bulgarian hazard regression model for first births we know that between 1989 and 1999 first conception risks have dropped by 49 in the first three years of a first partnership Using microsimulation we get the additional information that this change in parameters alone leads to a 48 higher proportion of childlessness an increase from 7 25 to 10 72 that it increases the average age at first birth by 9 6 months and decreases overall cohort fertility by 5 1 While such what if type studies are typical applications
41. rth on 2nd birth double Secon firstStatus FIRST STATE Relative risk union status at first birth on 2nd double Forml_age AGEINDEX Age baseline hazards first union formation double Forml_spellBirth BIRTHINDEX Rel risk time since last birth on lst formation double Form2_spellDisso DISSOINDEX Baseline time since dissolution of 2nd formation double Form2_spellBirth BIRTHINDEX Rel risk time since last birth on 2nd formation double Diss1_spellUnion UNIONINDEX Baseline time since formation on 1st dissulution double Diss1_spellBirth BIRTHINDEX Rel risk time since last birth on 1st dissolution double Diss2_spellUnion UNIONINDEX Baseline time since formation on 2st dissulution he parameter group P03 SECOND BIRTH aCe eSeCconde Barth wParamneters Secon age Secon_spellBirth Secon_unionStatus Secon _firstStatus he parameter group P04 PART FORM1 d First partnership formation parameters Forml_age Forml_spellBirth ha parameter group P05 PART FORM2 e Second partnership formation parameters Form2_spellDisso Form2_spellBirth he parameter group P06 PART DISS1 First partnership dissolution parameters Diss1_spellUnion Diss1_spellBirth he parameter group P06 PART DISS2 g Second partnership dissolution parameters Diss2_spellUnion he parameter group P07 DUR SPELLS h Definition of age and duration spells ageSchedule unionSchedule birthSc
42. s persons The characteristics and events changing the characteristics of actors are typically defined in other files that correspond to the various modeled behaviors In our application each person constitutes a case i e the whole life of each person is simulated before the simulation of the next person starts In more complex demographic applications birth events would result in the creation of new actors which then would belong to the parents case The number of simulated cases is part of the scenario definition and not set in the model code i e it can be set by the model user 7 The second time option is discrete time alternative model options are time based or cell based Currently no Modgen documentation is available for cell and time based models Modgen automatically creates two variables of the specified time_type time and age These variables are per default initialized to O and handled automatically in the simulation The initial value can be modified only in the Start method of an actor As we want to simulate a birth cohort born in the year 2000 we added the code time 2000 to the Start method The FinishQ method is applied at the end of life of an actor Actions to be performed before the actor is destroyed can be added to this method something that is useful e g to generate additional file output Before the simulation starts the Presimulation method is applied The method contains a SetMaxTime command whi
43. s account for 97 of all births reported in the population under study i e women of Bulgarian ethnicity born after 1950 Furthermore all first and second births in the sample occurred before a third partnership When modeling births we generally refer to the time of conception assumed to be nine months prior to the reported births This allows us to distinguish the period of pregnancy in models for union formation and dissolution inl x File Edit View Project Build Debug Tools Window Help H S sI rA Sh Debug Processes JIa gt eh dbarpt c clocks mpp lifecourse mpp M e Pg En Connect to Database t Connect to Server Main simulation file version 1 0 model_type case_based time_type double actor Person void Start void Finish he The Modgen tool to be run before building running the program Modgen a SEETEAEETEEALAEET ELE E LEA PEPLELEEEETEEETALAL ETAT Add Remove Toolbox Add in Manager Build Comment Web P Macros ActiveX Control Test q Create GUID Options The model code organized in mpp files Solution Explorer lifecourse a x S a C code generated by the The model parameter file The Modgen log Solution lifecourse 1 project J G lifecourse References MPP files clocks mpp lifecourse mpp mortality mpp parameters mpp Modgen to
44. se timing is not determined by the Monte Carlo simulation i e their timing is predetermined The only such event contained in the template model is a birthday event altering the integer age of the modeled actor Other clock events added later are events that change the spell index of the time varying covariates of the model The parameter definition file parameters mpp This file contains all definitions of parameters classifications partitions and ranges used in the model see below for explication For more complex models that have many different behaviors one may want to organize parameter definitions in a different manner e g by including them in the files of the corresponding behaviors The table definition file tables mpp This file contains the definition of output tables The actual model parameters are contained in one dat file Base lifecourse dat Again for more complex models it is advisable to split this file into one file for each group of modeled behaviors 2 2 2 The lifecourse mpp main simulation file This file contains the code essential for the definition of and for running case based continuous time Modgen applications The developers of these model types will need to modify only very few parts of the code which is mostly model independent At the core of any Modgen Simulation are the actors whose life is simulated In our application as in most demographic applications we have only one type of actor
45. tatus age entrances parity status ONE_CHILD entrances parity status TWO CHILDREN Av age at birth decimals 2 l1 entrances parity status ONE CHILD unit Childlessness decimals 4 entrances parity status ONE CHILD entrances parity status TWO CHILDREN unit Percent one child decimals 4 entrances parity status TWO CHILDREN unit Percent two children decimals 4 value at entrances life status NOT ALIVE parity unit Cohort Fertility Rate decimals 4 he table group TG02 FERTILITY TABLES De Ber talent 103 FertilityByAge T04 CohortFertility he 2 4 Finalizing the model Step 3 The third step finalizes our illustrative model The programming mainly consists in a repetition of tasks already explained i e the introduction of all parameters not yet introduced the definition of additional clocks second births and union transitions and the extension of table output 2 4 1 Finalizing the parameter definitions The behaviors introduced into the model at the final step are partnership and second birth Partnership transitions are parameterized by baseline hazards and control for different periods of motherhood Second conception is modeled accordingly the process starts at first conception and controls for age union status and partnership status at first birth In Step 3 we introduce three additional clocks see below This requires the definition of three additio
46. ting at second union formation and has no additional controls SECOND UNION DISSOLUTION lt 3 years union duration 3 9 years union duration 9 years union duration 0 04 0 01 0 07 31
47. traced both state variables namely the current age and the life status Figure 3 displays the life course for a person actor who lives 40 years see the BioBrowser User s Manual for details 12 t Modgen BioBrowser Biog1 loj x S File Edit Yiew Format Tools Filter Window Help plela Sla3 0 fam al 2 iter tracking 100 3 gt Current age gt Life Status 2000 2010 2020 2030 2040 Database C C Programs LifeCour Filter Person Ready Figure 3 The graphical BioBrowser output of individual life courses 2 3 Extending the model adding first births Step 2 In this the second step we extend the template by adding the first birth process First births are modeled by an age baseline hazard for different age groups and relative risks controlling for union status As we have not included union status yet for the present we assume a union status of never having been in union throughout life 2 3 1 Adding new parameters The key parameter concerning first birth is the age specific baseline hazard of the birth event First_age AGEINDEX Our model distinguishes 11 age periods 0 15 15 17 5 17 5 20 37 5 40 we therefore define a range AGEINDEX 0 10 of possible indexes We allow the model user to modify these age cutting points which we therefore also define as model parameters ageSchedule AGEINDEX The third parameter First_unionStatus UNION_STATE controls for unio
48. utting point in time 14 BULA ML AAA AEA ALAM AMAL AEA LALA ALLA Clock events file clocks mpp ADDITIONS STEP 2 Hi PUMA MAA LAAN LLAMA UNUUNU UUK UNUU A UUNGU LDA EMMA LA AAA LA MMA AMAL A AAA actor Person Individual int agespell 0 Current age group index TIME lastagespellchange TIME INFINITE Time of last age group index change event timeNextAgeSpellEvent NextAgeSpellEvent Event setting index of age group PA AAA EA AAA TEA AAALAC UUN UUU MUUMUU AUNA Event changing index of age group TIME Person timeNextAgeSpellEvent TIME event _time TIME INFINITE aioe ap for i MAX AGEINDEX i gt 0 i if age lt ageSchedule i event_time WAIT ageSchedule i age else if lastagespellchange time amp amp age ageSchedule i event_time time return event_time void Person NextAgeSpellEvent agespell lastagespellchange time 2 3 3 The event of first birth fertility mpp In order to keep the code for birth events in a separate file we create a new file fertility mpp as part of the project This can be done easily by adding a new text file item of this name to the list of mpp files via the context menu of the MPP files folder of the solution explorer Alternatively the file can be created outside the NET 03 environment and added to the project by choosing Add existing item from the context menu of the MPP files folder Aft
49. y dividing duration by unit As for parameter definitions comments placed into the code are used as labels in the application The second table is a period table Here we use the partition YEARS defined in parameters mpp to split up time into calendar years This is done by the Modgen split command The unit keyword now counts the entrances of actors into calendar year cells i e corresponds to actors alive at the beginning of a calendar year Accordingly the duration function returns the total years lived of all actors by calendar year which is equivalent to the average population alive in a calendar year As with the parameters tables can be grouped for easier navigation In our template model we define a table_group TGO1_LIFE_TABLES containing both tables 10 PUA ALA G ALAA AAA AAA AL ALATA LAA AA TA AAA AG Output tables tables mpp STEP 1 Hi FUGA EAA EAA AAA AAT AEA ALATA LAA LAD ALD LL LAA AAA ALLA ALAA ALLA table Person T01 LifeExpectancy 1 Life Expectancy UE Total simulated cases duration Total duration duration unit Average age at death decimals 3 he table Person T02 TotalPopulationByYear 2 Life table split time YEARS Output for each calendar year unit Population at start of year duration Average population in year he table group TGO1 LIFE TABLES a Life tables TO1 LifeExpectancy T02 TotalPopulationByYear he 2 2 7 Running th
Download Pdf Manuals
Related Search