I have used the
forecast::auto.arima() function to make an ARIMA model about a time series.
I have followed the great book by Rob J. Hyndman and George Athanasopoulos (2018) Forecasting: principles and practice, 2nd edition, OTexts: Melbourne, Australia. OTexts.com/fpp2. However, I am not sure about the results of my model, given that the distribution of residuals does not look much like a normal distribution, and that some of the inverse roots lie very close to the complex unit circle.
Any idea what might be happening?
Thanks in advance for any help!