I think eroding the region makes sense in general - you just have to be careful to exclude subjects who end up with no data (as you discovered!)
You could also look into why these three subjects had such small wm regions under entorhinal cortex. My (still) suspicious mind wonders if there is a bad reconstruction that is cutting out part of the inferior temporal lobe and therefore "squeezing" all of the cortical labels into a smaller volume. This would in turn give very small wm volumes.
On the other hand, entorhinal cortex may just have a relatively small amount of white matter underlying it because it's so "narrow" such that when you erode by 1mm you are left with nothing for these three subjects.
Glad you figured out the odd correlations!
Kx
Sent from my iPhone, please excuse any typos or excessive brevity
Hi Kirstie,
Yes indeed there were 3 subjects at the bottom-left of the plot.
To account for PV, I eroded the boundaries using -seg-erode 1 during the mri_segstats stage, and this resulted in a few subjects with no data for the WM entorhinal region (although these were saved as "0" in the .stats file). Do people generally do this for DTI studies?
All good now.
Many thanks again, Kirstie.
Best Wishes,
Elijah