Page 1 of 1

[Urgent] AR models in EViews9 uses all data points

Posted: Thu Aug 20, 2015 10:49 pm
by prit.ranjan
Hi,

Auto regressive (AR) model uses previous points residual as regressor. Logically a model with AR(1) term should not have any prediction for first point as AR(1) value will be unavailable for the first point. Similarly a model with AR(2) term would not have any prediction for first 2 points and so on. For some reason that is not the case with EViews 9.

Does anybody know how the predicted values are calculated for Auto regressive (AR) model in EViews 9? Attached are an example EViews 9 work file and corresponding excel calculation.

Thanks

Re: [Urgent] AR models in EViews9 uses all data points

Posted: Mon Aug 24, 2015 9:52 am
by EViews Glenn
EViews 9 defaults to using ML to estimate AR models. Like GLS, ML estimation uses all data points.

Dropping the first data points is done when you estimate with conditional least squares. On your options tab of your equation is a combo for choosing the ARMA estimation method. Select CLS.