I am curious if someone noted this aspect of SARIMAX class of statsmodels library and can share an opinion about the fact that it is not ignoring the seasonal parameters when periodicity is zero.
https://www.statsmodels.org/dev/generated/statsmodels.tsa.statespace.sarimax.SARIMAX.html
My understanding was that when periodicity is zero, than the Sarimax model should behave like a non-seasonal modeland ignore the seasonal parameters. But from the tests was clear that this was not the case.
Does someone knows why Sarimax is behaving like this? If I am not wrong, the R equivalent is ignoring the seasonal parameters when periodicity is zero.
Thanks
I tested (cross validation) a seasonal Sarimax model in two variants. In both tests, the 's' periodicity parameter was equal to zero and in one test P&Q seasonal params where different than zero while in the other were set to zero. The non-seasonal parameters were held the same in both tests. The result of the test (mape) was different in the two tests. I also used same exogen variables in both tests but this does not change the problem context.
My expectation is that Sarimax should ignore seasonal parameters when periodicity is zero.
