Autoregression when predicting yearly sum

[Taliah]

Hello

I need to build a statistical model using monthly data (each row is one month) predicting for each row the sum of variable Y for the next 12 months. I am using proc autoreg, and am getting Durbin Watson below or around 1 (i.e., autocorrelation in the data even after including AR variables). I understand why that is happening, as the model is predicting the sum of variable y for the following 12 rows. Any solution will be helpfull. Thank you.

[sbxkoenk]

I'm not sure I understand the full "depth" of your question and you haven't shared any code that could make me wiser.

But why don't you proceed with monthly data doing monthly forecasting (for the next 12 months ... lead=12)?

That yearly sum you need, can be done later in a post-processing data step. If you only need point forecasts, that may be the way-to-go. If you also need interval forecasts (confidence interval) then it becomes a bit more complicated.

Koen