Greetings FSL experts,
I've been running ANOVA on some VBM in randomise. I'm almost certain I've got the design and contrasts right based on Eugene's email of 27 May last (Re: Correct design setup for ANCOVA with nuisance variable).
So I've got 3 groups L, M, S and they are in the order of the ls command in the GM_mod_merg file.
The (abbreviated to 2 subjects per group) design and contrasts looks like this (I've been following the ANOVA: 1-factor 4-levels example on the webpage):
EV1 EV2 EV3
1 1 0
1 1 0
1 0 1
1 0 1
1 0 0
1 0 0
So assuming EV1 models the last group (S) the EVs are as follows
EV1: mean
EV2: L-S
EV3: M-S
Contrasts:
EV1 EV2 EV3 F-test
C1 L-M 0 1 0 *
C2 M-S 0 0 1 *
So assuming I've got this right (i'd be grateful if some one could confirm this for me), I then proceed to run randomise using TFCE, threshold (and binarise) the tfce_corrp_fstat1 and tfce_corrp_tstat files at 0.95 and use the resultant masks on to the corresponding fstat and tstat files. Is this correct should I be thresholding the tfce files like this and masking in this fashion? (The exact commands are below).
The problem I then have is that while I get regions showing a group effect from the fstat and some of these are shown in tstat1, the regions identified showing differences from tstat2 are not showing up in the thresholded fstat image. The fact that there regions showing in tstat2 that are not in fstat have me confused. Have I got something wrong with the method below or should non-overlap be expected?
Any insight on this would be gratefully appreciated.
randomise -i usersOnly_GM_mod_merg_s2 -o longVsMidVsShort -m GM_mask -d longVsMidVsShort.mat -t longVsMidVsShort.con -f longVsMidVsShort.fts -n 5000 -T
fslmaths longVsMidVsShort_tfce_corrp_fstat1 -thr 0.95 -bin mask/longVsMidVsShort_mask_pcorrected_fstat1
fslmaths longVsMidVsShort_fstat1 -mas mask/longVsMidVsShort_mask_pcorrected_fstat1 corrected/longVsMidVsShort_fstat1_corrected
for t in $( seq 1 1 2 ) ; do
if [ -f longVsMidVsShort_tfce_corrp_tstat${t}.nii.gz ] ; then
fslmaths longVsMidVsShort_tfce_corrp_tstat${t} -thr 0.95 -bin mask/longVsMidVsShort_mask_pcorrected_tstat$t
fslmaths longVsMidVsShort_tstat${t} -mas mask/longVsMidVsShort_mask_pcorrected_tstat$t corrected/longVsMidVsShort_tstat${t}_corrected
fi
done
Regards,
--
Colm G. Connolly, Ph. D.
Dept of Psychiatry
University of California, San Diego
9500 Gilman Dr. #0855
La Jolla, CA 92093-0855
Tel: +1-858-246-0608
Fax: +1-858-534-9450
|