Running into a peculiar problem with Forecast Studio.
If I subset the history to start from a later date than the earliest date in my data, Forecast Studio doesn't produce forecasts in some cases for a much longer historical period than the subset period.
E.g. I have a hierarchical dataset (forecasting by the hierarchy) of monthly data starting from Jan 2000 and going up to July 2003.
I don't want Forecast Studio to use data before say, June 2000.
When I tell Forecast Studio to do this (in the 'Data Preparation' tab), it produces forecasts for the historical data only from June 2001 -- ignoring all data points between June 2000 and June 2001 as well although I did not ask it to do this. These time periods have the history in them (Forecast Studio shows the history) but no forecasts.
However, it doesn't do this for all items in my dataset or at all levels in the hierarchy -- so, it correctly starts forecasting from the specified historical period in some cases but not in others.
Also, this missing historical forecasts issue seems to occur only if I move the 'Ignore data' pointer past a specific month (E.g. May 2000). So, even if I asked Forecast Studio to ignore data before April 2000, it correctly produces forecasts for the historical period. But, if I ask it to ignore before May 2000, then this problem occurs.
There's nothing strange about the period for which Forecast Studio refuses to produce forecasts (after I subset the date) -- the data is pretty much of the same pattern as the remaining time periods.
Any insights would be greatly appreciated!
Please take a look at the selected model for the series with less hsitorical fits than the historical data. ARIMA models with any degree of diffing and lagging will consume a few data points at the beginning of the historical periods, and thus no fitted values will be generated. Series with full historical fitted values are generally results from ESM (smoothing) models or ARIMA with no diffing/lagging.
Alex
Please take a look at the selected model for the series with less hsitorical fits than the historical data. ARIMA models with any degree of diffing and lagging will consume a few data points at the beginning of the historical periods, and thus no fitted values will be generated. Series with full historical fitted values are generally results from ESM (smoothing) models or ARIMA with no diffing/lagging.
Alex
Join us for SAS Innovate 2025, our biggest and most exciting global event of the year, in Orlando, FL, from May 6-9. Sign up by March 14 for just $795.
Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.
Find more tutorials on the SAS Users YouTube channel.