For different country data to be useful to users, the model must be at least crudely calibrated to that setting, which the user can then refine. When the user loads a region, it should load not just the data but the parameter values for that region. These calibrations should be performed by us offline, to the level of fidelity we have time for, and saved as a library of parameter values in the webapp folder.
[ ] Perform manual calibration to ~5 locations, determining which parameters will need to be varied (e.g. beta, initial infected population, change_beta date and magnitude)
[ ] Choose calibration algorithm
[ ] Apply and test calibration algorithm to match deaths data
[ ] Once deaths are calibrated, calibrate optimal symp_prob value in cv.test_prob() to match diagnoses
[ ] Verify calibrations for ~20 locations (~5 US states, ~5 high-income countries, ~10 low-income countries)
For different country data to be useful to users, the model must be at least crudely calibrated to that setting, which the user can then refine. When the user loads a region, it should load not just the data but the parameter values for that region. These calibrations should be performed by us offline, to the level of fidelity we have time for, and saved as a library of parameter values in the webapp folder.
symp_probvalue incv.test_prob()to match diagnoses