Dear Mark,
> I am trying to put a covariate into an SPM analysis. The covariate is
> based on visual analogue scales, and I suspect that different subjects
> have used different levels on the VAS to mean the same degree of the
> variable I am trying to measure. I therefore wonder if I shouldn't be
> using "specific fit" for this covariate. Is this a reasonable thing to
> do?
Very reasonable (and more comprehensive than simply mean correcting the
subject-specific VAS scores before entering them as a single
covariate).
> Secondly when I put the covariate in as "specific fit" it then asks for
> 11 items in each contrast vector. (I assume because I have 11
> subjects). I entered 2 contrasts, 1 1 1 1 1 1 1 1 1 1 1 and -1 -1 -1
> -1 -1 -1 -1 -1 -1 -1 -1, again is this reasonable?
Yes it is. These contrasts will give you the average regression slopes
over subjects. Note you can also test for differences among subjects
(i.e. subject x VAS interactions) with the appropriate contrasts using
this model.
Bets wishes - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|