I am trying to replicate the results from proc AUTOREG in Excel. I have no problem replicating the forecasting (out of the sample) values. I can't however obtain the same results as SAS for the estimation (in sample) part. I am sure it's due to the way AUTOREG computes the errors for estimation, which seems to be completely different than for the forecasting part. There is no documentation available about this. Can you help me figure out this problem?