Page 1 of 1

NLS versus simple OLS

Posted: Tue Sep 12, 2017 10:58 am
by oksanakim
Hello

I have a couple of simple questions regarding model specifications. It seems my EViews version was updated to no. 10 (I think) and I noticed some differences in the models spec. When I enter the model spec in the Equation window and click on the Estimation settings window, I see "LS - Least Squares (NLS and ARMA) as the first option. I believe there used to be simple OLS in the place of NLS in the previous versions, correct? So how do I now estimated simple OLS model?

And second, when would one want to use ROBUSTLS option from the Settings list? Thanks very much!!

Re: NLS versus simple OLS

Posted: Tue Sep 12, 2017 11:07 am
by EViews Gareth
OLS is covered by the LS option.

If you specify a non-linear model, NLS will be used. If you specify a linear model, OLS will be used.

Re: NLS versus simple OLS

Posted: Tue Sep 12, 2017 11:09 am
by oksanakim
Thank you! How about ROBUSTLS option? When would this one be used? In the case of a linear model, what is the advantage of using this option?

Re: NLS versus simple OLS

Posted: Tue Sep 12, 2017 12:35 pm
by oksanakim
And a related question: in the Binary estimation settings, how do I know when to use "extreme values" model and if it would be preferred to Probit and Logit specifications. Thank you.