I have a set of dependent variables which are serially correlated. I understand the confidence and prediction intervals are actually much broader than the calculated intervals. If I cannot eliminate the autocorrelation, is there a way to determine the actual prediction intervals?
↧