 |
 |
Hi Anderson, Thank you for your reply! I used the following command line to run randomise: randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 --T2 -D. I'll send the design files to your email address instead of the mailing list. Thank you, Judith From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Anderson M. Winkler Sent: Freitag, 13. Juni 2014 13:24 To: [log in to unmask] Subject: Re: [FSL] Again: Demeaning in TBSS Hi Judith, This sounds weird. Which version of FSL are you using? Could you please send the exact design.mat and design.con files (for both with and without manual mean-centering)? Also, please, paste here the exact command line(s) you used when running randomise. Thanks! All the best, Anderson On 13 June 2014 09:12, Judith Daniels <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear Colleagues, I processed my DTI data with TBSS. Now I want to run a correlation between FA and a psychological variable in a single group. I set up a very simple design: EV1 (group): all 1s EV2 (covariate) My contrasts were 0 1 (for pos. correlation 0 -1 (for neg. correlation) I ran randomise without demeaning the data as advised in Janette Mumford's guide to mean-centering, example 1 (http://mumford.fmripower.org/mean_centering/) and got highly significant (albeit unlikely) results. I re-ran the same design and contrast using the -D option after manually demeaning my covariate (6 decimals included). In my understanding doing this should lead to the exact same output. However, now my results are not significant at all anymore. I then extracted FA values for all subjects from the peak voxel and ran simple correlations in SPSS between FA and demeaned and un-demeaned covariate values as well as demeaned_FA values and demeaned and un-demeaned covariate values. As expected, I always get the exact same result in SPSS. So why do my FSL results change so dramatically and which result is the correct one? Does using the TFCE option have to do with this? Thanks for any suggestions you might have, Judith