search

CDCgov / wastewater-informed-covid-forecasting

Wastewater-informed COVID-19 forecasting models submitted to the COVID-19 Forecast Hub
https://cdcgov.github.io/wastewater-informed-covid-forecasting/
Apache License 2.0
44 stars 8 forks source link

Modify eta and inf feedback #198

Closed kaitejohnson closed 3 weeks ago

kaitejohnson commented 3 weeks ago

This is based on the logmean and logsd of the posteriors run for 365 days for the 5 example states, figures here https://github.com/cdcent/cfa-forecast-renewal-ww/issues/774.

These posteriors from the above PR were used to estimate the infection feedback and R(t) RW step size, acknowledging that we are now using our training dataset on our testing data set (fitting to the fill forecast period and then comparing performance), basically tuning the model.

These are logmean = log(90), logsd = log(1.89) for infection feedback. For eta_sd the mean was 0.278, which for now we will use as the sd (can change so mode is not 0 if this works well)

kaitejohnson commented 3 weeks ago

Plots of results compared to previous priors we have tried, green is this latest run. I think the overall results are pretty driven by the forecasts from 2024-01-08, so the relative rankings might change if we were to have run all the forecast dates for all locations.

subset_forecasts_by_forecast_date subset_forecasts_by_location subset_forecasts_overall

I am going to go ahead and do another run where I change the prior mode of eta_sd in the wwinference stan model, since this was just changing the prior on the stdev in the RW step size with a mode still at 0.