FSL experts,

- I apologize, this is a resend/bump as I didn't hear back and could really use some help! -

I am attempting to relate changes in resting connectivity in healthy seniors and AD patients to behavioral covariates of interest while "regressing out" the effects of others. I'd like to use randomise to eventually do permutation testing, but I'm having trouble getting my contrast design sussed out properly in the GLM setup. Effectively, I'm trying to replicate the methods of Bijsterbosch et al., 2013 (J. Cog. Neurosci), but looking at pain related indicies (behaviors/ratings) rather than anxiety related questionnaires.

I think I have a decent handle on how to set up the design.mat file of EVs. However, and this may be a stupid question, but do my specific correlation coefficients go into this part of the GLM setup?

Most of my confusion pertains to the contrast design bit, though. Let's say I have two groups, 10 subjects in each, with 3 covariates of non-interest (age, depression scores, and gender). These are followed by several covariates of interest, namely pain ratings and behavioral scores for pain responses. All covariates are to be mean-centered.

Here is a sample of EVs, with all covariates of NON interest, and one OF interest, just to simplify things...

HS AD Age Gender Pain_rating1

1 0 1.97 -.29 3.4

1 0 -8.03 -.29 .4

1 0 -5.03 -.290 .4

1 0 3.97 .710 .4

1 0 -3.03 -.290 -2.35

1 0 -8.03 -.290 -4.6

1 0 -2.03 -.290 -.6

1 0 2.97 .710 -3.1

1 0 -4.03 .710 1.4

1 0 5.97 -.290 -.1

0.0 1.0 -1.03 -.290 4.4

0.0 1.0 -5.03 -.290 2.9

0.0 1.0 12.97 -.290 -1.1

0.0 1.0 -5.03 .710 -1.35

0.0 1.0 -6.03 -.290 1.4

0.0 1.0 5.97 -.290 1.9

0.0 1.0 2.97 -.290 2.4

0.0 1.0 2.97 -.290 -.1

0.0 1.0 9.97 .710 .4

0.0 1.0 2.97 .710 -2.6

To look at group differences, adjusted for all covariates, the contrasts would be:

HS>AD: [1 -1 0 0]

HS<AD: [-1 1 0 0]

Now, if I want to look at the relationship of connectivity differences on behavior, how would I set up my contrast?

Also, how about if I wanted to look at the relationship between connectivity differences and TWO behavioral covariates of interest?

Thanks so much for your help!

--
Paul Beach
DO/PhD candidate
Michigan State University
College of Osteopathic Medicine
- Medical Scientist Training Program
Neuroscience Program