Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Weird MFSDA result compare to original Matlab. #274

Open
iPsych opened this issue Jun 30, 2022 · 0 comments
Open

Weird MFSDA result compare to original Matlab. #274

iPsych opened this issue Jun 30, 2022 · 0 comments

Comments

@iPsych
Copy link

iPsych commented Jun 30, 2022

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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant