Hi Andy,
Judging by the error messages, it seems that your X matrix doesn't
have full rank. I would check on this and drop the variable(s) that
are causing the linear-dependency problem from the model.
Hope this helps.
Best,
Kevin
------------------------------------------------------
Kevin Quinn
Assistant Professor
Department of Government and
Center for Basic Research in the Social Sciences
34 Kirkland Street
Harvard University
Cambridge, MA 02138
On Tue, 28 Dec 2004, Andy Harris wrote:
Hi All,
I'm working with some household level data I collected last summer, and
have run into some heteroskedasticity problems. I've tried to correct
for it both using White's correction and through a robust estimation
technique, but neither work. I get the following errors:
for Whites:
whitese<-sqrt(diag(hccm(lm2)))
Error in
V %*% t(X) : non-conformable arguments
for rlm:
Error in rlm.default(x, y, weights, method = method, wt.method =
wt.method, :
x is singular: singular fits are not implemented in rlm
All but one of the independent variables are dummies.
Any clues on how I can either get around these errors, or correct for
the heteroskedasticity in another way?
andy
_______________________________________________
gov1000-list mailing list
gov1000-list(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov1000-list