- What are some basic tips on going back to the drawing-board after a seasonal ARIMA fit shows significant coefficients (1,0,0)x(0,1,0) but fails in its Ljung-Box stats?
- Can minitab be used to model an ARIMA with regression?
Look at the autocorrelation and partial autocorrelation graph of the residuals and see what sort of peaks you see. This may suggest what other terms might be needed. The answer may be "none". If you are just seeing resigual autocorrelation at either long lags or unusual lags without any way to explain them (e.g. lag 9 on monthly data), you may want to ignore the Ljung-Box failure.
Minitab's ARIMA won't go beyond the type of model you are fitting.