EViews.com

Hello All

I am looking for the way to estimate an OLS model using the Rogers robust standard errors, which is supposed to control not only for firm-level dependence, but also for dependence within each cluster. I understand that White correction corrects for the former but not for the latter. Is there any built in option for the Rogers se? I haven't located any and the Eviews guidance does not contain any discussion for this. Thank you.

My understanding is that Rogers robust standard errors are a name for clustered standard errors. I'm not sure why we need yet another name, but there you have it. Clustered standard errors are available in EViews in a panel structured workfile and are denoted White-period standard errors (since we conceptually think of there being cross-period, within cluster correlation).

There are many, many threads in the forum on estimating clustered standard errors. A quick search brings up, for example:

http://forums.eviews.com/viewtopic.php? ... ster#p4315

What you'll probably need to do is to structure your workfile as an undated panel with cross-section ID given by the cluster variable. Once you do, you'll have access to the panel structure covariances including White-period. But that's just an assumption; if you describe your data in greater detail someone can offer additional guidance.

Thank you Glen, I will try your suggestions. Much appreciated.