growth_rate has some error

[dsweber2]

@dajmcdon mentioned there was some kind of error in the `growth rate calculation

[dajmcdon]

I believe only the method = "rel_change" will work correctly. The rest should not be used. A good stop gap would be to just remove those as possible arguments. But eventually, we'd want to fit the growth rate correctly on training data, and extrapolate correctly for test data.

[dsweber2]

just confirming, is the problem here or in epiprocess? If I used test_data that was the full data of the training set, would it work correctly? I'm thinking about using this as a replacement for most of the derivative estimates in a flusion-like model, as the way I wrote those is very slow.

[dajmcdon]

just epipredict.

Yes, the error is that the calculation made in get_test_data only applies when method = "rel_change". If you give it the whole training set, then it's fine.

[dajmcdon]

I think I removed the other methods for the moment?

And the "proper fix" in my mind is to deprecate get_test_data in favor of a fix to forecast() such that it takes the whole training set, appends any new data, does all the preprocessing, then returns only the predictions at the test data (rather than selecting a subset of data needed to do the preprocessing).