[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Robust Standard Errors
From:       Jim <jlinck () UGA ! EDU ! NOSPAM>
Date:       2002-10-31 21:33:35
[Download RAW message or body]

Suppose one has a data set with a possible independence problem (for
example, say a panel of information on people tracked for several years).
It's logical to assume that the observations across years for each person
are not independent.  Thus, robust standard errors to account for
"clustering" within a person would be necessary when testing the
significance of individual coefficients.  However, does this hold for F-test
across coefficients as well?  That is, suppose I have:
 Y = X1 + X2 + X3 + X4
and I was concerned with independence of my observations.  If I wanted to
test:
 X1=X2
 X3=X4
(X1-X2)=(X3-X4)

Could I do F-tests using a standard OLS model for these tests?  On a related
matter, what if I assume there is an idependence problem, but their really
isn't?  Would this reduce my power?

Any thoughts would be appreciated? Thanks!

[prev in list] [next in list] [prev in thread] [next in thread]