Dear SPM team,
I am running a VBM analysis in a cohort of subjects before and after they have received a treatment (no controls).
We have ran fluid registrations in those subjects to evaluate if there is voxel-wise volume increase/decreases over time and want to show which brain areas show significant volume change over time after adjusting for covariates. We also want to check if changes correlate with certain clinical scores.
In terms of setting up the matrix for the longitudinal change I was planning to set up the matrix as below:
age sex site1 site2 site3 site4
32 1 1 0 0 0
24 0 1 0 0 0
33 0 0 1 0 0
53 0 0 1 0 0
45 1 0 0 1 0
55 1 0 0 1 0
25 1 0 0 0 1
40 1 0 0 0 1
And the following contrasts (1 or -1 to see if there is significant change followed by as many zeros as covariates)
1 0 0 0 0 0
-1 0 0 0 0 0
For the correlations however I would set up the matrix as below
clinical score age sex site1 site2 site3 site4
25 32 1 1 0 0 0
33 24 0 1 0 0 0
40 33 0 0 1 0 0
22 53 0 0 1 0 0
10 45 1 0 0 1 0
15 55 1 0 0 1 0
17 25 1 0 0 0 1
29 40 1 0 0 0 1
and the contrast like this (zero for the intercept, 1/-1 to see if there is positive/negative correlation and one zero for each covariate):
0 1 0 0 0 0 0 0
0 -1 0 0 0 0 0 0
My questions are as follows:
1. Do the design matrix and contrasts look ok?
2. Should I demean each variable, including the binary variables, before entering them into the design matrix?
3. Is it ok to add as many variables as sites -1 and code them as 0/1 instead of 1,2,3,4?
4. I have excluded also two sites with only one subject in each of them.
Sorry for the long email and multiple questions, just wanted to ensure everything is ok and easier to understand for the expert!
Many thanks,
Best regards,
Carlos
|