notes/stable_tmb.md

In bbolker/McMasterPandemic: Pandemic Model

Potentially Relevant Changes

• https://github.com/mac-theobio/McMasterPandemic/commit/b3973afe2ad859254adcea1d700d08d569c4e9f9
• Expandify
• make_vaccination_model
• https://github.com/mac-theobio/McMasterPandemic/commit/d7cbbed680896ae03d9e8ea89a73954846659211
• Common multiples
  • get_var_indx, get_var_counts, get_var_modifiers
• Eigenvector tolerance
• Spec version bump to 0.1.2
• (NO -- never used) #include <cppad/local/cond_exp.hpp>
• https://github.com/mac-theobio/McMasterPandemic/commit/36d5bcd5602b6a0b7cea1a5f5f7893a039f50061
• getOption("MP_tmb_models_match_r")
• https://github.com/mac-theobio/McMasterPandemic/commit/dc2b2ccb1238c040e5a289d029d4235f66b1feb8
• winstretch -- almost surely not the issue
• https://github.com/mac-theobio/McMasterPandemic/commit/b0f1c6c5082f35f07bbf50522a5a0cc9bc54a498
• nt <- round(gg[["Gbar"]] * winstretch) -- this rounding could be the issue, but probably not

Hypotheses

• make_vaccination_model change
• this is the most likely issue
• can be tested by r_tmb_comparable()
• evec tolerance bug
• this seems like the most likely to me
• causes difference in initial state
• might be able to prove a difference by comparing initial TMB-computed state on both branches
• factrs
• seems very unlikely given that make_vaccination_model does not include the factr feature -- this has only been used for threedosewane
• also this feature has been tested reasonably well -- although never in production
• expandify
• likely not the issue, unless expand_params_variant is called by the pipeline
• but even then ...
• winstretch rounding
• perhaps this causes differences in fix_pars, but not sure
• can mike test this?

single arrow from R_fully_vax back to S_fully_vax --

bbolker/McMasterPandemic documentation built on Aug. 25, 2024, 6:35 p.m.