Back to PLS HelpPosted on 05/12/16 22:05:18
Number of posts: 16
Hi all,
I was wondering the best way to quantify how similar the "activation maps" of two LVs are.
Is it reasonable to simply correlate the saliences (or bootstrap ratios) across all voxels, with higher correlations indicating higher similarities?
Regards,
Reece
Replies:
Posted on 05/13/16 06:56:33
Number of posts: 394
Hi all,
I was wondering the best way to quantify how similar the "activation maps" of two LVs are.
Is it reasonable to simply correlate the saliences (or bootstrap ratios) across all voxels, with higher correlations indicating higher similarities?
Regards,
Reece
you can correlate the saliences or if the LVs are unit length still (sums of squares =1) you can compute the dot product which is the cosine of the angle aka correlation
the one caveat is that you should ensure the length of the vectors is the same. The PLS code extracts "brain" voxels from the image deck using a masking routine (captured in the 'coords" matrix) and if the masks are not the same, the correlation will take a bit more housework to calculate. Let me know if the coords matrix is the same between analyses
Login to reply to this topic.
