Todd,
On Thu, 8 Jun 2006, [log in to unmask] wrote:
> These residuals are Yhat = Y - X betahat were "hat" are the
> estimates of the respective variable?
In general, yes, that's right. Though in this limited setting, we're
thinking of X as being just the drift basis, right?
> What I'm asking is if these residuals are in fact the estimates of
> the errors? If so, then the residuals are the K(s).X0*(K(s).X0'*y)
> in the expression on line 172 in spm_filter?
No. K(s).X0'*y is the projection of the data unto the drift basis,
and corresponds to "betahat" above, and then K(s).X0*(K(s).X0'*y) is
akin to "X betahat", the estimated drift effect. Once the estimated
drift effect is subtracted off from Y, *then* you have residuals.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|