Hello,   I have a Forecast Studio project with a weekly time scale and thus the seasonal cycle length is 52. I've noticed that when I add an event to a copy of an automatically generated Subset ARIMA model that has DIF = (1 52), the event is then assigned DIF = (53). All of the other events that have been automatically included in the model have the same differencing as the model, i.e. DIF = (1 52). To the best of my knowledge, an event's differencing should be the same as the model's differencing. Is there an explanation as to why it assigns DIF = (53) instead of (1 52)?   Thanks, Blake ... View more

Hello,   My group creates monthly forecasts using SAS Forecast Server (not sure how to tell what version), and does manual checks using the Java Web Launcher version of Forecast Studio (14.1). For one month in particular, we create a forecast at three different times in the year. For this month's forecasts, we like to keep all of our model selections and re-estimate the parameters using the "FIT" option for %fsrunprj()'s method from one checking period to the next:     %fslogin(user=user_name, password=pass_word, sasEnvironment=default); %fsrunprj(projectname=project, method=FIT, importdata=yes, environment=environment); %fslogout();     For other month-to-month occasions, we use the "DIAGNOSE" method at the start of each checking period after adding new historical data points, and I have never seen the issue below.   Occasionally, our selection from the previous month will be rejected after refitting. Personally, I've only ever seen this happen to an ARIMA model (not combined or ESM). Again, in my experience, there is no reason why this model should be rejected. In all of the cases that I've found, making a copy of the rejected model will fix the issue, and the parameters will be successfully evaluated for the original model.   Is there any reason why this is happening? Is there any way to avoid it other than the costly process of making the copy model?   Rejected model is highlighted. The model itself was a copy of a system generated ARIMA from a previous forecast:   Copy has been created (and removed for simplicity), which fixes the issue:   Thank you, Blake ... View more