ARIMA model, tools, diagnostic problem

Question

Describe the diagnostic tools you would use to determine if your model is appropriate! What can we learn from these tools? Why do we care? Which tests are used to detect the diagnostic problems in a time series analysis?

Answer 1

Look at the autocorrelation, partial autocorrelation, and extended autocorrelation. If model a moving average, there is a cut off in the autocorrelation function. If model autoregressive model there is a cutoff in the partial autocorrelation function. If model is a mixture of moving average and autoregressive, extended autocorrelation will provide you information . Look at the value of the AIC or BIC calculated when building your model. The best fit will have the lowest AIC or BIC. When building the models, use all the time the same information criteria, i.e, all the time AIC or all the time BIC.