Hi Gang
Even-though the filtering is performed in advance to GLM estimation, the
information is carried through. To convince yourself try to perform the
analysis with and without highpass filtering. At the bottom of the results
page you can see the degrees left, this will be lower for the analysis
without high pass filter.
Best
Torben
> Hi,
>
> I understand that high-pass filtering is a prerequisite for auto-
regressive modeling for individual subject analysis and is done
> through spm_filter.m in SPM2. It seems to me that the filtering
> process is done before the modeling; in other words, those lower
frequencies are removed beforehand, and are not included in the final
design matrix (and thus not accounted for in the final effective degrees
of freedom). However on the SPM website it states the "AR(1)
characterisation is only appropriate when long-term correlations due to
drift have been modelled deterministically through high-pass
> filtering or, equivalently by inclusion in the design matrix as
> nuisance variables." Could somebody kindly clarify my confusion?
>
> Thanks,
> Gang
>
|