Validate the behavior of MultiPCA if a MultiNiftiMasker is given about nilearn HOT 8 CLOSED

After discussion with Philippe, we find much more consistant to ignore masker
parameters given to MultiPCA in this case (if the user gives us a masker, we
assume that he knows what he's doing).

I'd rather have it the other way: explicit is better than implicit.

from nilearn.

Reverted.

from nilearn.

My thoughts on this.

MultiPCA needs a mask for its computations. The user can provide it or let the object compute it from data. The latter situation can be handled in two ways: either by providing a masker object, or by providing the options taken by the masker object (and let MultiPCA create one). What the user would write is respectively (in extremely simplified form, for the purpose of the example):

MultiPCA(smoothing_fwhm=2)

or

masker = MultiNiftiMasker(smoothing_fwhm=2)
MultiPCA(mask=masker)

The second situation being useful when the same masker (not mask) is to be reused in different places.

The question raised in this issue is what should happen in this situation:

masker = MultiNiftiMasker(smoothing_fwhm=2)
MultiPCA(mask=masker, smoothing_fwhm=3)

What I think after talking with Alexandre is that this situation must not happen: this is an error. The Python langage prevents a keyword argument from being provided twice. Let's do the same. The present example is a conflict, which is ambiguous. Giving priority to either value (that in masker or that provided to MultiPCA itself) is implicit in all cases.

from nilearn.

For the moment, if we are in such a case, the value of the MultiPCA is taken and we raise a warning. If it's OK, then we can close this issue.

from nilearn.

For the moment, if we are in such a case, the value of the MultiPCA is taken
and we raise a warning.

I am not sure what your sentence mean.

from nilearn.

if there is a conflict between the parameters of the MultiNiftiMasker and of the MultiPCA, the parameters of the MultiPCA are kept and a warning message tell the user that parameters of the masker were overridden

from nilearn.

if there is a conflict between the parameters of the MultiNiftiMasker and of
the MultiPCA, the parameters of the MultiPCA are kept and a warning message
tell the user that parameters of the masker were overridden

All good!

from nilearn.

OK, I close this issue then.

from nilearn.