Sample 90% of the observation as the training set, then apply forward and backward selection approaches; Compare out-of-sample MSPE between these reduced models.
go to top