 |
 |
Hi, This case is covered under "Two-Group Difference Adjusted for Covariate" in the GLM section of the FSL wiki. It doesn't matter that the covariate is gender. The way you've specified things in your email isn't the usual way of accounting for gender - where we would normally just add a single covariate that contained a demeaned set of values indicating gender (e.g. +1 and -1 if the groups were balanced, or +0.5 and -0.5 as the magnitude is arbitrary). However, as long as you are using correct contrasts then you can use your 4 EV version if you prefer. You should not put zeros in the columns related to gender if you just want to account for (or correct for) gender. So contrasts of 1 -1 0 0 and -1 1 0 0 would be what you want for the 4 EV matrix model. If you used a 3 EV model instead (where the third column was a demeaned version of gender) then just drop the last entry in these contrast vectors. All the best, Mark On 11 Sep 2014, at 06:45, Victor DelBene <[log in to unmask]> wrote: > Dear FSL experts, > > In TBSS, I would like to run randomise using gender as a covariate. I'm having some confusion regarding the design.con file and how that needs to be set up. To provide context, I included an example of my design.mat file. > > This is an example of my current design.mat setup (without all 79 participants included). > > /NumWaves 4 > /NumPoints 79 > /PPheights 1 1 > /Matrix > > EV1 (Controls) EV2 (Patients) EV3 (Males) EV4 (Females > 1 0 1 0 > 1 0 1 0 > 1 0 0 1 > 1 0 0 1 > 0 1 0 1 > 0 1 1 0 > 0 1 1 0 > 0 1 0 1 > > In my design.con file, I currently have: > /NumWaves 4 > /NumContrasts 2 > /PPheights 1 1 > /Matrix > > 1 -1 > -1 1 > > The design.mat and design.con file were created when I first ran TBSS without covariates. I want to now see if my findings hold when I account for gender, so in the design.mat file I added EV3 and EV4 (as shown above). Since the NumWaves is now 4 in the.mat and .con files, I believe I need to have something like this in the design.con file: > > EV1 EV2 EV3 EV4 > 1 -1 1 1 > -1 1 -1 -1 > > Would this be the correct way to format the.mat and the .con file to run randomise with two groups and a covariate of gender? > > My apologies if this has been asked before, but in researching this on the mailing list, the FSL Wiki page on randomise, or tutorials from other labs, I haven't really found a sufficiently clear answer to my question. > > As always, thank you for your help! > > Best, > Victor