04-11-2012 10:26 AM
Hi, has anyone ever done multiple time series in a regression model.. So for example if I have 3 different regions (Prov) each with 10 sub-regions (like City) and each sub-serie each with 132 months... How can I create a regression model for each of the 3 regions but have the model behave (or predict) individually by the 10 sub-regions? To add to the problem, I have variables that some are at the Region level and some that are at the Sub-region level.. If I want the sub-region variables to be used, do I leave it at the region * sub-region level?.. I know tim-series (ARIMA) will do the trick, but either way do I have to build 3*10 or 30 models?