Hi Laurent,

That is a good point and you are exactly right that for single-subject statistics using Fisher's test to compare two seed-to-voxel r-maps it is more appropriate to use the "after denoising" approximation to the residual degrees of freedom (this takes into account not only the correction in degrees of freedom due to band-pass filtering operation but also one due to the regression of potential confounding effects). I will make the changes to the code to have the DOF variable in those mat files store the proper/improved values, and thanks again for pointing this out 

Best
Alfonso

ps. as a note mostly for future reference, compared to population-level (multiple-subject) statistics, single-subject statistics like these are always complicated by the degree to which the stats assumptions are truly held in our data; in this case the underlying assumptions would be having relatively white/colorless residuals in the BOLD signal over the frequency window of interest after denoising, and having relatively large number of samples and not-too-high correlations; perhaps other methods that carry less assumptions, such as phase-randomization for example, could be used in the future to produce more accurate/robust single-subject statistics when needed

Originally posted by laurentcohen: