Hello,
So I redid the analysis and this is the proper way to do the analysis, Right?
1- 3dDeconvolve using the -errts option
2- 3dFWHMx -automask -input subj_errts
3- Average the FWHM values across all subjects which yielded 8.6
4- 3dClustSim -mask gr_14subj_mask_NoRd+tlrc. -dxyz 3 3 3 -fwhm 8.6
and all this yielded a cluster size of > 39 for p=0.005 (voxel) and p=0.05 (cluster)
Say that I'm not all that happy with the output, alternatively, I could use FDR correction but I'm not sure how to use it.
In the t-test I have there is q values but they 'saturate' at 0.27 and the corresponding p value of the t-test is 3e-4, which is pretty low. Does it mean there is a probability of false positive of 0.27% ? That does not sound very good!!!
Also, do I have to implement a cluster analysis and if so how or do I just live with whatever is present after thresholding the stats map with q values? I mean which p value would use as input to 3dClustSim? ??
Thank you again for your help
pat