signaturescience / focustools

Forecasting COVID-19 in the US
https://signaturescience.github.io/focustools/
GNU General Public License v3.0
0 stars 0 forks source link

retrieve parameters from ARIMA models and store during pipeline #48

Closed vpnagraj closed 3 years ago

vpnagraj commented 3 years ago

at the moment when we run ARIMA models through the ts_* functions we use an auto ARIMA process to choose P,D,Q parameters. we constrain the parameters to a given space, but the values selected may change week-to-week.

we should include some code to retrieve the parameters and store as a variable. from there we could save those values to an object and write out during forecast pipeline.

the trick will be trying to not break our existing code. is it possible to get the parameters out of the object returned by ts_fit()? or would we need to crack into that function and change the output to a list or something?

@stephenturner assigning to you for now but i'm happy to help as needed.

vpnagraj commented 3 years ago

btw this might be related to https://github.com/signaturescience/focustools/issues/17

meaning ... if we can figure out the attr method for data could we add some kind of attr for ARIMA PDQ without changing the downstream behavior of the object coming out of ts_fit() ?