Ensure all lagged variables are either set by input data at T=1 or set to null if missing

[andrewbaxter439]

From discussion at development meeting on the 28th Jan:

• Ideally, lagged values should not be imputed when missing
• If an initial population doesn't have lags, these should not be necessary
  • all updating processes should only use yearlySchedule.addCollectionEvent (as these are not updated in year 1
• If methods incorrectly require lagged variables they should crash
• Check logic for children coming of age and immigrants to ensure correct handling of null lags

[justin-ven]

New tags added to improve attribute initialisation (pull request #453 ). Specifically, tags @lag and @NullInitialised have been added to the project to facilitate initialisation of attributes at the start of each simulated period. These tags have not been added to candidate attributes where the model was found to crash or generate different results. The issue is most accute for the Person objects, where tags have been omitted from the following candiates:

• yMiscPersGrossMonth
• yCapitalPersMonth
• yPensPersGrossMonth
• yEmpPersGrossMonth
• yNonBenPersGrossMonthL1
• yNonBenPersGrossMonth
• labWageFullTimeHrly
• labHrsWorkWeekL1
• labHrsWorkWeek;
• demPartnerNYearL1
• xEquivYear
• labWageFullTimeHrly
• labHrsWorkEnumWeek

[Mariia-Var]

All the attributes that lack @lag or @NullInitialised tags, (e.g., some of them are tagged as enumerated or transient) aren’t re-initialised at the beginning of the year, so by default they carry forward the previous period values.

I have attached an excel file summarising all Person and BU class attributes, together with possible candidates for @NullInitialised and @lag. The document can be used as a guide to check the attributes that are currently carried forward.

person_attribute_lifecycle.xlsx