Weird MFSDA result compare to original Matlab.

[iPsych]

Dear @bpaniagua and @vicory,

I experienced a bit weird experience with slicersalt implication (actually, same in the original python implication too) of MFSDA. Below is the detail.

For example, Let's say testing 2 groups with 2 additional linear covariates. Total 3 covariates in the model. When run MFSDA via below python script, it gives 3 clu_pvals no matter how many clusters are formed, even nothing survives in Lpvals_fdr.

/Applications/SlicerSALT_Latest.app/Contents/MacOS/SlicerSALT --no-main-window --python-script /Applications/SlicerSALT_Latest.app/Contents/lib/SlicerSALT-4.11/qt-scripted-modules/MFSDA_run.py

When test with matlab version, the original one, https://github.com/BIG-S2/MFSDA the number of clu_pvals varies depend on the number of formed clusters. -single cluster is formed : 1 x 3 -two clusters are formed : 2 x 3

I guess something gone wrong in numpy, so the python implication saves the 1st value (0th in array) for each covariate. I am checking the python code now, but please look into.