reporting exact, FDR corrected p-values for results obtained with GLMFlex/FIVE

[Jan Beucke]

Hi Aaron, 

first of all thanks so much for sharing GLMFlex, I am working with twin data which involves analysis of repeated measures in several groups, so I am very grateful to have this tool! 

I have one question: How can I report exact, FDR corrected p values in FIVE?

When I use FDR correction in FIVE, I can use different thresholds (e.g. p < .05, or p < .01), but I can not see the exact p -value (e.g. p = .02) which is required by some journals. The exact p-value displayed in the GUI must be the uncorrected p-value, as it does not correspond to the FDR correction thresholds that I apply.

Thanks so much for your help!

Best,

Jan

[Aaron Schultz]

In the menu bar, under Parameters->Corrected Alpha, you can set the
desired alpha level for the correction. The p-value/t-value listed in
main interface is the uncorrected threshold needed to achieve a
corrected alpha at the desired level (0.05 by default).

-Aaron

> --
> You received this message because you are subscribed to the Google Groups "Aaron's fMRI matlab tools." group.
> To unsubscribe from this group and stop receiving emails from it, send an email to fmri_matlab_to...@googlegroups.com.
> To post to this group, send email to fmri_mat...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/9f560b4e-5b1b-4fcf-8199-27a1f5794c93%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Jan Beucke]

Hi Aaron,

thanks for the quick response! 

I understand that, but is there a way to report the exact corrected p-value (instead of the level and/or the uncorrected p-value corresponding to it?), the way it is done in SPM results Tables for example, where the corrected and the uncorrected p-values are provided?

Best,

Jan

To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAH2ncsvDp70PnERjSO8wGWTF1xaomOYtEy9FsbhN5xV0WZAa7w%40mail.gmail.com.

[Aaron Schultz]

I'm not sure what you are referring to. What other p-value would
there be? Are you referring to cluster statistics?

-Aaron
On Tue, Oct 2, 2018 at 10:28 AM 'Jan Beucke' via Aaron's fMRI matlab

> To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAKQiQSUvKAoyB%3Dyftf%3Db1PZyZvmy5S1eXzr7wmV6-zYU2Cm_ZA%40mail.gmail.com.

[Jan Beucke]

Hi Aaron,

I am attaching a screenshot of table example from the SPM manual, which provides exact p values for results at cluster- and peak-levels, respectively ( P FWE and P FDR values), in addition to uncorrected P values on the right side of the table.

So I would like to know how I can get that exact peak-level P FDR value from FIVE.

Please don't get me wrong, in my view the P FDR thresholding option that is provided in FIVE is totally sufficient, it is just that some journals insist that exact p-values should be reported, and thus I would like to know how I can manage not only to provide the correction threshold that was achieved, but also the exact corrected P FDR value.

Again, thanks for your help with this!

Jan

To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAH2ncst3FA2NujNX1yByt2u4ZYvcrLtaXUt-JC2vkr0ZMc_rvA%40mail.gmail.com.

[Aaron Schultz]

What you want is probably peak_nii, which will work with FIVE. Once
peak_nii is installed and set up, you can do Options->Get Peak Info.

-Aaron
On Tue, Oct 2, 2018 at 10:52 AM 'Jan Beucke' via Aaron's fMRI matlab

> To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAKQiQSWXnqv_Bm47E4x2XLDyihkAH4fpt7brp-8r_OEUYLYrPA%40mail.gmail.com.

[Jan Beucke]

I actually tried that before, but it gives me the following error message:

Reference to non-existent field 'FIVE'.

Error in peak_nii (line 120)

FIVEF=mapparameters.FIVE;

Error in FIVE/getPeakInfo (line 3247)

        [peak.voxels, peak.voxelstats, peak.clusterstats,

        peak.sigthresh, peak.regions, peak.mapparameters,

        peak.UID] = peak_nii(Obj(vn).FullPath(1:end-2),S);

 

Error while evaluating Menu Callback

>> 

To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAH2ncss%3Dg8uotAEdPys-1wbmk-nETJHmVfZ%3DLJG3UfAMz12qKA%40mail.gmail.com.

[Aaron Schultz]

I would follow up with Donald McLaren on troubleshooting peak_nii.

-Aaron
On Tue, Oct 2, 2018 at 11:26 AM 'Jan Beucke' via Aaron's fMRI matlab

> To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/CAKQiQSUztyFYKTmposLH6QTAvs_GePmWRJAKzqDKDmPJBKeM7w%40mail.gmail.com.