Dear Brandall,
> In a second level analysis I am getting essentially the same results
> with and without the covariate in the analysis. Also, the
> cluster-level corrected p-values are in some cases lower than the
> voxel-level uncorrected p-values. How can this be?
This is fine. Remember that the uncorrected and corrected p values
based on peak height pertain to that voxel (i.e. the probability of
obtaining a cluster of more than zero volume). The probability of
getting a cluster of finite and large volume is much less and can be
less than the uncorrected p value.
> In the first level analysis I am outputting contrast images from two
> sets of conditions for each of two gender groups. In the second level I
> am comparing the two groups (as advised) by means of a "Single-subject:
> conditions & covariates" analysis -as one subject, two conditions
> (groups as conditions) and one covariate. Is there some fault in this
> way of analysis?
This seems perfectly OK.
I hope this helps - Karl
|