01-27-2015 02:05 AM
I have a question in regards to regression analysis in EG. More pacifically when ti comes to Stepwise Selection in regression analysis. What I am referring to is the ability to specify a data set to be used as an output to the model. I know that Stepwise section goes through a lot of inputs and determine the best model. And you can specify additional data set to apply the results of the regression to in order to determine Predicted outcome.
As well there is comprehensive log that comes with the regression which shows how many steps were preformed in order to accomplish the task. But nowhere does it point to the exact Step (model) that SAS EG chose itself in order to fill in the projected outcome.
Now it is easy to find such model when you have 5 or 6 steps to examine, but when it come to 60 or so steps it is not so easy. Especial when AIC, BIC, SBC, R squered and C(p) don't all show the same step as the best possible model.
I would relay appropriate if someone would be able to provide me wit a bit more information on this issue, or if you need a bit more explanation I would be more than happy to elaborate further.