data. Will prpbably need tweaking.

guessing the data type of a file/directory path.

used.

VoxelwiseEV and VoxelwiseConfoundEV classes share a common base class.

this is all handled by nibabel. Also never keeps its own ref to file object.
Image.save always saves to a temp file, then moves that to the destination -
avoid any problems with mmap/indexed_gzip.

reduce the number of series that need to be loaded.

in file.

get image data as numpy array.

affines

state. Image.save ensures that file name has an extension. MGHImage uses both
of these changes so that its initial save state is accurate, and when it
is saved, it is saved as a nifti

check data shape, and raise if bad.