Bring `aaa_fit_multi-sexbehav-sae` up to speed with `aaa_fit_all-dhs-multi-sexbehav-sae`

athowes / multi-agyw

Spatio-temporal estimates of HIV risk group proportions for AGYW across 13 priority countries in sub-Saharan Africa

https://athowes.github.io/multi-agyw

MIT License

2 stars 2 forks source link

Bring `aaa_fit_multi-sexbehav-sae` up to speed with `aaa_fit_all-dhs-multi-sexbehav-sae` #36

Closed athowes closed 2 years ago

athowes commented 2 years ago

I think there are improvements I've made to the later which haven't been included in the first. That said, the difference between the two scripts is mainly 4 categories versus 3 categories -- since for all-dhs I've included an if else loop saying that if there is only one survey then don't bother with the temporal random effects.

athowes commented 2 years ago

Ah, the reason this could be important is that I need the 4 category outputs to compare to Laga in #8. So would prefer that they are good quality.

athowes commented 2 years ago

The utility script for running this should only run for countries with the good FSW question

athowes commented 2 years ago

Included country level aggregates
Used update for the formulas to save a bit of writing
Cleared up confusion about using BYM2 here or Besag (we're using Besag!)
Removed the part about max_model_id (not a great idea, there are only three models and they fit quickly)
Updated the multinomial_model function code to have the country level aggregates etc.