Hey Rick,

Ah, this will help a ton. Figured niml might be the solution, but wasn't aware of adding FDR curves with 3drefit.

I realized there will be one other challenge with FDR curves for my specific application, and curious if you or anyone may have any thoughts:

My analysis involves parcellating high-resolution standardized surface datasets (163842 vertices) into 1280 regions per hemisphere. The data I am analyzing pertain to each region (I average the original signal within each region).

Therefore, FDR is a challenge because I have to map (using Matlab, e.g.) 1D files with the 1280 datapoints to "viewing" datasets of 163842 - so that all vertices belonging to one region have the same data. Therefore I have the data in 1D format until converting to "viewing" datasets, but adding FDR curves at the end would cause afni/suma to do multiple comparisons for 163842 rather than 1280 observations.

The easiest solution to this I can think of for q-values is just to stick with the '1-q' approach because I will probably just use a standard FDR of 0.05, or 0.01. But there should be no problem with the uncorrected p-values, which will now appear dynamically from the T-stat/F-stat data, so this is still an improvement over before. However, I'd be curious if you think there's a solution to this issue - perhaps the FDR curves can be tweaked, etc. I may play around with this a bit when I get to this point, and will post any success. Otherwise, will use niml dsets for p-values and my crude '1-q' thresholding for q-values.

Thanks!

John