trislett/TFCE_mediation

Question about results summary

ajdneuro12 opened this issue · 3 comments

ajdneuro12 commented

Dear Tristram,

I had a question about my results summary (see attached). I was wondering about the minimum vertex size for a cluster to 'count'. Is there some sort of threshold for TFCE. All of the clusters are significant at .05 FWE but some are 8 vertices or 2 vertices. Is there some sort of cutoff that you recommend. Thanks for the great work on the toolbox!

Best,
Alex
lh_thickness_results_summary.txt

trislett commented

Hi Alex,

Sorry for the late reply.

There's no cut-off since the vertices are already FWER corrected. So when you have a 'cluster' that is only 2 or 8 vertices, it is likely that many of their neighbours are slightly subthreshold or it just doesn't connect with a larger cluster. I recommend using using the clusters only to plot the mean values from on the top areas (i.e., cluster 1 or cluster 2 in your case). tm-tools has an option for extracted the mean values.
In my opinion, it it much more useful to visualize the results using freeview or TMI_viewer.

Cheers,

Tris

ajdneuro12 commented

Hi Tris,

Thanks for your response. I am writing up this paper now but I am curious for the results summary which would you report in the paper, all of the clusters? Which would you report in a table? Thanks!

Best,
Alex

trislett commented

Hi Alex,

For my own preference, I would report the FWER corrected image as figures for the main result. This best represents the actual statistical analyses performed.

Cheers,

Tris