Yikes.
We have dutifully search the archives, but I haven't been able to find a clear example
of...
..... the appropriate matrix for modeling a subject-related covariate of interest (e.g., a
mood score, etc.) on a second level contrast for multiple subjects in a randomized,
event-related fMRI study.
Our design:
Each subject experienced 4 distinct categories (A, B, C, D) of stimuli, plus nulls.
For the first-level, each subject has 4 con files: one for each type of stimulus category
vs. the null : e.g., A vs. null, B vs. null, C vs. null and D vs. null.
These four con file types for each subject are the basis for the second-level of the
design, which is to compare A vs. D, B vs. D, C vs. D (where D is a "neutral" stimulus,
and the others have different affective valences).
To the covariate question:
I would like to incoporate a mood score (taken at time of fMRI session, for each
subject) as a covariate of interest; the score may model the variance differently for
each the three second-level contrasts (A vs. D, B vs. D, and C vs. D).
[[[[When we assign the same mood covariate to each of the four rows representing A,
B, C, and D (for a given subject), and -- for example -- then select a (A vs. D, etc. )
contrast plus the column containing the mood score covariate for "A" ....SPM gives
the message "invalid contrast"....so the model we have selected doesn't accept the
same (subject mood) covariate 'attached' to each of the conditions.]]]]]
Most grateful for any guidance.
[[Significant dark chocolate dessert to be mailed UPS to the clearest responder!!]]
Anna Rose Childress, Ph.D.
Neuroimaging Division
Addiction Treatment Research Center
Department of Psychiatry
University of Pennsylvania School of Medicine
3900 Chestnut Street
Philadelphia, PA. 19104-6178
|