IBMA estimators resampling input (potentially non-optimally)

[jdkent]

Summary

when preprocessing the input for IBMA, the input images are resampled to match the masker object (through the transform).
nilearn's resampling function default options may not be optimal for meta-analytic maps.

For example, some code from @tyarkoni:
resample_to_img(nb.load(img), template, clip=True, interpolation='linear')

deviates from the defaults interpolation='continuous' and clip=False.
Changing those defaults appears to give the maps being analyzed a more biologically plausible result.

Additional details

• NiMARE version: dev

What were you trying to do?

add nimare to the neuroscout paper

What did you expect to happen?

biologically plausible results conformant to previous analyses with PyMare

What actually happened?

weird pixelated results