Hi everybody,
I'm trying to conduct a VBM study on two groups of subjects. It has taken us years to acquire the data. In that time we've had to contend with scanner replacement so not only have I two groups I've also got two scanners. My groups are split almost evening across the scanners.
Can anyone provide any advice on what precautions I should take to minimize the effect of scanner?
I've conducted a preliminary analysis using randomise where I included the scanner as a covariate of no interest. I coded the scanner as 1 and 2 in the matrix file.
Is it sufficient to code the scanners as 1 and two or should they have been coded as -0.5 and +0.5, i.e. demeaned. Note that in this case I ran randomise as follows:
randomise -i GM_mod_merg_s2 -m GM_mask -o sVcScanner -d sVcScanner.mat -t sVcScanner.con -T -n 5000 -V
The matrix indued one column for group s, another for group c, and a third for the scanner (this latter one was coded as 1 or 2 depending on scanner). The contrast matrix was as follows
1.000000e+00 -1.000000e+00 0.000000e+00
-1.000000e+00 1.000000e+00 0.000000e+00
Any advice would be most greatly appreciated.
Regards,
--
Colm G. Connolly, Ph. D.
Dept of Psychiatry
University of California, San Diego
9500 Gilman Dr. #0738
La Jolla, CA 92093-0738
Tel: +1-858-246-0684
|