Hi Marlene,
The AFD values are stored as fixel data files, and do not in isolation contain any spatial or orientation information; so a dimension mismatch between such a file and a 3D mask image is inevitable.
What you are looking for is a fixel mask: A fixel data file that contains a value of 1 for fixels that you want to include in your computation, and a value of 0 for fixels that you do not want to include. There are various ways that such a mask could be derived; but the obvious starting point is the voxel2fixel command, which will project values from a 3D voxel image into the fixels underlying each voxel. You may however wish to refine this selection further, since a voxel-based selection will include unwanted crossing-fibre fixels in your selection; admittedly the tools for doing this are still a little lacking (e.g. editing fixel ROIs within mrview is technically feasible), but the combination of tck2fixel and mrthreshold can be useful.
From there, it’s simply a matter of executing mrstats on the AFD fixel data file, providing your derived fixel mask using the -mask option.
Rob