Sorry if this posts twice. I got an undelivered message from my email
service.
Dear All,
Does anyone know why SPM high-pass filters both rows and columns of
(W*V*W') when estimating the pre-whitened variance. In (e.g. Woolridge
2001) the variance estimate is given by:
var = ResSS / trace(R*W*V*W');
but SPM does the filtering before calculating the above trace (which
does give different results) so I'm curious as to what the accepted
convention is and why SPM chooses to filter before calculating the
trace.
Regards,
Brian
|