I tried doing minimal changes to use a non-deprecated implementation. Works, but there is a problem: the results I obtain are different. See the test... Is this expected?
If we can not solve this is quickly I suggest pinning statsmodels <= 0.13 and making a release.
@markotoplak, your changes seem OK. I tried to reconstruct the history of changes in ARIMA to check that your code is equivalent to what we had before ... but I'm giving up. I say we merge this.
Issue
177
Description of changes
I tried doing minimal changes to use a non-deprecated implementation. Works, but there is a problem: the results I obtain are different. See the test... Is this expected?
If we can not solve this is quickly I suggest pinning statsmodels <= 0.13 and making a release.
Includes