Dear FSL-Experts,
I have a question concerning my TBSS-Analysis executed with randomise (no –D option, no meancentering) in a group of patients versus healthy controls.
I want to perform a FA-Group Comparison and a within group correlation for Co-Variates of interest and testing for interactions of these Co-Variates.
Therefore I have the TBSS-Data and a bunch of other EVs which I want to correct for in the FA-Group Comparison (Co-variates of no interest) and also I want to experiment on with Correlation and Interactions (Co-Variates of Interest).
I know how to set up the contrasts for my purpose, but I have a question about the EVs in my Design, which I will explain in the following:
The Matrix in general has the following design (as an example with Age, Gender and Rating Scale):
G1 G2 Age(G1)Age(G2)Gen(G1) Gen(G2) RS(G1) RS(G2)
1 0 78 0 2 0 11 0
1 0 65 0 2 0 24 0
1 0 73 0 1 0 3 0
0 1 0 78 0 1 0 13
0 1 0 68 0 1 0 25
0 1 0 60 0 2 0 12
In My Actual Study I Do Not Only Have Age Gender And One Rating Scale. I Also Have Two More Rating Scales, Two Different Cortisol-Data
AND
fMRI-data to nine brain regions acquired under 4 different neuropsychological conditions (but in the same MRI session). To sum it up: 36 different fMRI-Co-variates of no Interest, each of which will get two EV’s as showed in the matrix above (same goes for the two cortisol data and the other two rating scales).
So far i did my group comparison (and Interaction-Analysis) with 5 different Designs:
1. corrected for: Age Gender Cortisol and the rating scales
2. corrected for: Age Gender and the fMRI-first condition
3. corrected for: Age Gender and the fMRI-second condition
4. corrected for: Age Gender and the fMRI-third condition
5. corrected for: Age Gender and the fMRI-fourth condition
But is it also possible to put everything in one design-matrix instead of splitting it up in these 5 Designmatrices? Or would that be to many Co-Variates of no interest, which i want to correct for? Or would that affect or confound the statistics?
In case it works to put all the Co-Variates in one Matrix for Group comparison, can I use the same matrix (with a different contrast file of course) for the correlation and interaction-Analysis?
I hope you can help me with this issue!
I appreciate everything you can comment on!
Kind regards
Berel Weinhöppel
Department of Neurology, Universitätsmedizin Mannheim
University of Heidelberg
68167 Mannheim, Germany
|