It depends on a couple of things. If you don't have any continuous covariates, then the difference will be in the computation of the noise and the DOF. With all groups, you'll be computing the noise variance across all subjects, and you'll have a much higher DOF. The size of the difference between group1 and 2 (ie, gamma.mgh) should not change in the two designs. If you do have continuous covariates, then things get more complicated.

doug

Alex Millner wrote:

Hi -

Thanks for your help.

I have 3 groups and am trying to run an ANOVA and series of t-tests. First, I ran mri_glmfit with two rows in my .mat file (1 -1 0, 1 0 -1) for the ANOVA. Then, before I ran into this webpage https://surfer.nmr.mgh.harvard.edu/fswiki/FsgdExamples, to do a follow up t-test between group 1 and group 2, I tried to mri_glmfit with a mat file of 1 -1 but it gave me an error because it was expecting any test to include all the subjects. So reran isxconcat-sess with subjects from just groups 1 and 2, and ran mri_glmfit with a mat file of 1 -1

Then I found the FSGD Examples webpage and realized I could perform all the tests in one command as long as I included a 0 in my mat file. So for the contrast between groups 1 and 2, I made a mat/mtx file with 1 -1 0.

My question is, why are my results from these two similar analyses very different from each other? There are many more activations in the 1 -1 0 version than in the 1 -1 version and the activations that do exist are, for the most part, in different areas of the brain. Sorry to be ignorant of some of the statistics/analysis differences that are taking place here. I wondered why it was different and which was a more valid analysis. It seems like doing one analysis with all the tests is a better way to go then divided the data up, but I just wanted to check.

Thanks again for your help, Alex

Below are the mri_glmfit commands I used

Directory PDSubsONLY = contains output of isxconcat-sess run on placebo and drug subjects only

Directoy Analysis_Results = contains output of isxconcat-sess run all 3 subjects groups

mri_glmfit --y PDSubsONLY_Results/RSF_v_NIF/tal.ces.nii --fsgd FSGD_Placebo_v_Drug.txt --C Placebo_v_Drug.mat --mask PDSubsONLY_Results/tal.mask.nii --glmdir Analysis_Results/tal.wrfx.RSF_v_NIF --nii --wls PDSubsONLY_Results/RSF_v_NIF/tal.cesvar.nii

mri_glmfit --y Analysis_Results/RSF_v_NIF/tal.ces.nii --fsgd FSGD_Placebo_v_Drug_v_Control.txt --C mtx_files/placebo-drug.mtx --C mtx_files/placebo-control.mtx --C mtx_files/drug-control.mtx --C mtx_files/group_effect.mtx --C mt x_files/placebo-drug+control.mtx --C mtx_files/placebo+drug-control.mtx --mask Analysis_Results/tal.mask.nii --glmdir Analysis_Results/tal.wrfx.RSF_v_NIF --nii --wls Analysis_Results/RSF_v_NIF/tal.cesvar.nii

Placebo_v_Drug.mat = 1 -1

mtx_files/placebo-drug.mtx = 1 -1 0

FSGD_Placebo_v_Drug.txt = contains only placebo and drug subjects

FSGD_Placebo_v_Drug_v_Control.txt = contains placebo, drug and control subjects _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer