[Question] Permutation blocks for HHG

[hausler89]

Hi!

Is there a reason why the perm_blocks parameter of IndependenceTest.test() is not captured by the derived method of most of the independence tests, like HHG().test()? Can HHG or Hsic not be used for block-independent samples?

Cheers!

[sampan501]

We haven't tested it in other methods, which is why I haven't implemented it yet

[hausler89]

I see, thank you. But in principle, it should work if the block permutation is an accurate representation of the null, right?

[sampan501]

It should work, but I don't have any simulations prepared giving insights on power comparisons

[hausler89]

I will do simulations for the specifics of my project to see if HHG works for me. I'll see how I proceed from here. I'm probably wrapping your HHG implementation into my own little permutation loop; I think I need a TQDM progress bar for my sanity anyway.

Thanks for the help!