The weekly time series consistently looks too conservative, presumably because of the crude 1-week truncation. However the time series are always used and can be up to 100% of the ensemble weight in some states.
Probably best fixed by adding daily time series forecasts - the code is there, just needs integration.
is the time series truncated because there is evidence in the data of truncation? If yes we should do something about it more generally as it will bias all of the other models (EpiNow2 etc) as well.
The weekly time series consistently looks too conservative, presumably because of the crude 1-week truncation. However the time series are always used and can be up to 100% of the ensemble weight in some states.
Probably best fixed by adding daily time series forecasts - the code is there, just needs integration.