I can answer part of your question, I think. If I'm wrong, at least might generate a response from others on the list for that reason.
I would include your variable of interest with the rest of your covariates/nuisance variables and not as an interaction (are you doing an ANCOVA- analysis of covariance- for your 2nd level analysis?). I don't know about centering.
As far as generating the contrasts:
1) "For each degenerative group we'd like to determine where our covariate of interest correlates with gray matter tissue that is not already explained by the atrophy due to disease"
Would you agree if I rephrased this as: BOLD response for each group that is accounted for by the covariate of interest alone? If so, enter 1 as the contrast weight for the group and covariate of interest, -1 for control group, 0's for everything else.
2) "Regions correlated with the covariate across groups, again not explained by atrophy vs. controls"
If you want to figure out only those brain regions correlated with the covariate of interest that all 4 groups have in common, you will probably have to use a combination of masks of the results of the 4 analyses (from above). Maybe someone else can chime in here because I'm no expert on masking.
Hope this helps.
Julie E. McEntee, M.A., C.C.R.P.
Senior Research Program Coordinator
Department of Psychiatry- Neuroimaging
Johns Hopkins University School of Medicine
600 N. Wolfe St./ Phipps 300
Baltimore, MD 21287
Phone: 410-502-0468
Fax: 410-614-3676
----- Original Message -----
From: Rich Crawford <[log in to unmask]>
Date: Monday, November 26, 2007 11:00 am
Subject: [SPM] Contrast design
To: [log in to unmask]
> Dear SPMers,
>
> I haven't heard from any list members and I am still unsure how to proceed.
> Any input would be greatly appreciated.
>
> Our SPM2 model consists of four conditions, one for control subjects
> and the
> other three each represent a different neurodegenerative disorder.
> We have
> a single covariate of interest, two nuisance covariates and a global
> correction factor.
>
> For each degenerative group we’d like to determine where our
> covariate of
> interest correlates with gray matter tissue that is not already
> explained by
> the atrophy due to disease. We would also like to see if there are regions
> correlated with the covariate across groups, again not explained by atrophy
> vs. controls. Is it correct to model our covariate of interest as an
> interaction? Does centering matter? And how would we generate contrasts
> to answer the two questions above?
>
> Many thanks,
>
> Rich Crawford
>
> UCSF Memory and Aging Center
|