[Gretl-users] Dealing with heteroskedasticity