Open JohannesWiesner opened 1 year ago
Slow reply, apologies. I think the above is entirely possible and I obviously mothballed the project to focus on other priorities.
I agree with the principle though and I think in some ways it was a design problem of the MATLAB PALM package that it wasn't particularly modular so pulling bits out was tough!
In fact I am going to archive this repo to avoid confusion. My mistake fwiw was trying to replicate PALM when I should have tried to rebuild the useful functionality in a modular (pythonic) way. Hopefully hierarch are achieving that :)
Describe the bug
Hi James, I would like to use
skpalm.permutations.quickperms
to create permuted ids that respect the covariance structure of the HCP. I always assumed that for the resampled arrays we would draw with replacement but in such a way that that the covariance structures in the resampled samples match the original covariance structure. But right now, I just end up with shuffled subject is (for which I could just have usednp.random.shuffle(subject_ids)
. My current guess is that I must have understood something wrong?To Reproduce
BTW, I think it already mentioned the hierach package, very nice python implementation for hierarchical resampling. Right now, the package is still tied to very specific tests and thus not usable for CCA but I already asked the maintainer to separate the resampling methods from the statistical tests that one wants to conduct.