[Gretl-users] Out-of-sample forecasting with cross section data