Determine whether COVIDCast jhu-csse adequately represents the state data used by Reich

cmu-delphi / covidcast-indicators

Back end for producing indicators and loading them into the COVIDcast API.

https://cmu-delphi.github.io/delphi-epidata/api/covidcast.html

MIT License

12 stars 17 forks source link

Determine whether COVIDCast jhu-csse adequately represents the state data used by Reich #350

Open krivard opened 3 years ago

krivard commented 3 years ago

Ok, so I'm still not sure what the right answer is then. We want to predict the "truth". If the truth is XX000, then great. But if XX000 \= JHU's province_state, and we don't have JHU's province_state signal, then we need a fix. I don't believe that we actually need the bins separately. (@bnaras or @capolitsch can be more definitive)

[...] long term, wrapped up in some of the anomaly detection or data checking, we should make sure that our truth and their truth is the same (unless Delphi doesn't believe their truth)

Originally posted by @dajmcdon in https://github.com/cmu-delphi/covidcast-indicators/pull/328#issuecomment-713782551

dshemetov commented 3 years ago

@dajmcdon Just want to clarify that regardless of how we solve the JHU Out of State and Unassigned state signal binning, our state geo level signal should match the Reich Lab's state geo level signal. Our county level XX000 bin is essentially a granular "probably and miscellaneous" counts bin that Reich lab does not report. But in terms reproducing state-level truth, our JHU state signal should be identical to Reich Lab's. We can double check this once JHU is healthy again.