Dear Dough,

Thanks for your reply. Below, I am summarizing my concerns from previous emails:

Q1. I was wondering which of the following two options is better than the other (or neither may be !):

(i) CFT < 0.001 and CWT < 0.1 (i.e., CWT = 0.07) (smoothing 12 mm) for cortical thickness-behavior analysis (using MCZ)

or (ii) CFT < 0.05 and CWT < 0.001 (i.e., CWT = 0.0002) (smoothing 10 mm) for cortical volume-behavior (using MCZ)

Q2. I tried 3rd option using PALM (most recommended !). I found that one of the two clusters I found is very identical (visually) between PALM (CFT = 0.05) and MCZ (CFT = 0.05 and CWT = 0.001) for cortical volume -behavior association at smoothing 12 mm. However, the second cluster survives with MCZ (CFT = 0.05 and CWT = 0.001), but not for PALM (CFT = 0.05). I have following three concerns:

(a). Could you please confirm if PALM (CFT = 0.05, 10,000 permutations) with following command is the best approach here (compared to MCZ)?

palm -i y.mgh -s fsaverage/surf/lh.white fsaverage/surf/lh.white.avg.area.mgh -d X.csv -t C1.csv -m mask.mgh -o plm -C 1.95996 -Cstat extent -twotail -n 10000

If that's correct, then to save summary result file from PALM, I ran mri_binarize, mris_calc and mri_surfcluster commands (by looking at previous discussions in FreeSurfer forum). My concern is that the PALM summary file gives one big cluster with peak within the orbitofrontal gyrus, but visually it shows that it extends to fusiform and temporal regions. Could you help me in figuring out the way to generate summary file in such a way that I get summary of each region (using Desikan atlas) where ever results extend (as opposed to the one I am getting now i.e., based on where the peak is, and because there is only one peak so I get summary of one big cluster around that peak)?

(b). The MCZ results clearly tell whether the correlation is negative or positive (by looking at the cache*.summary and cache*.cluster.mgh files). But I could not find a way from PALM results to make sure the correlation is negative or positive. All the outputs e.g., *.fwep.mgz and *.tstat.mgz show positive maps always. Could you help in finding the directionality of correlations from PALM results?

(c). For MCZ I provide flag --2spaces (for Bonf. Correction of hemispheres) so I assume that the above PALM command does not perform Bonf Correction, and if so then I applied Bonf. Correction in mri_surfcluster command using the flag --bonferroni 2. Could you please confirm is that's the correct way?

Thanks.

On Fri, Jan 10, 2020 at 10:43 AM Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edu> wrote:

sorry, I'm not sure I'm following. It looks like you did two analyses, one with 10mm smoothing and CFT=.05, the other with 12mm and CFT=.001. You get a cluster for each in the same area, but they are not overlapping. Is that right?

On 1/6/2020 2:58 PM, Martin Juneja wrote:
External Email - Use Caution

Hi,

I am using Monte-Carlo simulations (for cortical thickness and volume - behavioral analysis) for clusterwise correction for multiple comparisons.

My results are either significant at (i) CFT < 0.001 and CWT < 0.1 (i.e., CWT = 0.07) (smoothing 12 mm) for cortical thickness-behavior analysis - with maxima at -4.3 (CFT)

or (ii) CFT < 0.05 and CWT < 0.001 (i.e., CWT = 0.0002) (smoothing 10 mm) for cortical volume-behavior - with maxima at -4.53 (CFT)

Both the clusters in (i) (lateral orbital frontal) and (ii) (rostral middle) are within the frontal cortex, but not overlapping !

Also, although maxima is at (i) -4.3 and (ii) -4.53, still the clusters do not survive recommended thresholds suggested in https://www.ncbi.nlm.nih.gov/pubmed/29288131

And the FPR in this paper at strong CWT (i..e, 0.001) and liberal CFT (i..e, 0.05) is not calculated.

So I was wondering if there is any recommendation from the experts regarding these results, or none of these make sense because I never get CFT < 0.001 and CWT < 0.05 (at-least).

Any help would be really appreciated.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer