I am currently following the Udemy lecture for time series analysis link.
When running pmdarima 1.7.1 auto_arima (statsmodels 0.11) on a pretty basic dataset, I am receiving a summary that just has the model stating SARIMAX with no p,q,d. See image below.
Should I just take this as the model being all 0s or white noise, since the 'aic' is stating 823.489, which traces back to ARIMA(0,0,0)(0,0,0)[0] intercept?
When running in an older version of pmdarima (1.10) and older version of statsmodels (0.9), I receive different results. See below.
Does the newer version of auto_arima not report out on ARMA anymore, and I should just reference the coefficient for the general linear mean?
I am only putting a few arguments in the auto_arima function currently.
auto_arima(df1['Births'],seasonal=False,trace=True).summary()
After digging in pmdarima versioning I found that with version 1.5.1 of pmdarima, this function will not longer use stats model ARMA and ARIMA. It will now only SARIMAX.
Change logs for this library can be found below.
Python pmdarima auto_arima Newest Version Issue
I guess my question now, is will auto_arima still give accurate forecast for stationary/non seasonal data?