Hi Afni Experts,
I have used freesurfer smoothed images (lh.thickness.fsaverage.fwhm15.gii) to perform a cortical thickness analysis. I used 3dttest++ to compare two groups. As covariates of no interest I added age (years), sex (0/1), site(1-10) and years of education(years). I used the following command:
3dttest++ -prefix Group1vsGroup2_Lthickness15 -covariate Group1vsGroup2 -AminusB -setA Group1 ..DSETS.... -setB Group2 ...DSETS
When using no covariate file and thresholding to p=.05 (t=2) in SUMA and setting tmin=2 in qdec and comparing these results of those in Freesurfer's qdec the results look very similar (matching previous posts).
1) However, when using covariates my results were very different from Freesurfer. I am trying to figure out the differences in how covariates are treated in 3dttest++ as compared with qdec.
--In this case should centering diff and pooled variance be the default if I did not explicitly specify?
--If i just have site 1 through 10 listed as a nuisance covariate to account for site differnces does it make sense to use these options?
--For age/years of education does centering diff seem reasonable?
2) Do you know why running through ttest++ without covariates look almost identical to freesurfer but running with covariates looks completely different? I am not sure if anyone has looked into this already.
3) Should we use some sort of FDR or clusterwise extent for correction with the cortical thickness? I looked into slow_surf.py for FDR but im not sure which program to use in order to do a clusterwise correction. Once I run these programs how do I apply it to my results? As an input I used fsaverage_lh spec and fsaverage_SurfVol for slow_surf.py Do i need to add my ttest output into this command somehow?
Thanks,
Ajay