Open cmhoove14 opened 3 years ago
First step is just to write scripts to assess outcomes produced by model simulations. data/outputs/Test_Run_bta...
contains an .rds file with the simulation outputs:
epi_curve
contains number of agents in each step at each time stepinfections
contains entries for every infection eventlineliest_tests
contains all tests conductedSo epi curve can be used for general calibration (e.g. to hospitalizations), but finer calibration by race and/or by geographies best done using infections
and linelist_tests
. Some ideas on how to calibrate/what we're looking for:
ct
) with low healthy places index/high minority race/low incomeCan we avoid race-based testing probabilities? I think it would be a stronger analysis to identify if/when racial disparities arise organically via calibration to known data, rather than hard code racial inequity into simulation.
@sblumberg Good point, and I agree. income already affects testing probability and now that synthetic population includes the healthy places index, have a few other variables that could be used (believe there's a healthcare access score, for instance)
Calibration underway for general transmission dynamics, but also want to calibrate to health disparities by race or at least geography.
Goal is to approximately match case and/or hospitalization rate(s) by racial/geographic group
Parameters to tweak could be race-based testing probabilities
Analysis/02-Prep-Par-Inputs ; input_pars$test_pars$race_test_mults
. Nothing else in model is explicitly tied to race, so might have to try deeper tweaks too