Back to PLS HelpPosted on 11/07/14 06:06:51
Number of posts: 13
Dear PLS experts,
I ran PLS analysis with two regions (not a single voxel) as seeds. Each seed region included 27 voxels. I got two significant LVs, but I want to exclude the seeds (where the correlation between variables in X and Y are perfect) from the result. Could you please guide me to be able to exclude the seed regions? Since this could have been a common problem for many users, I hope a code is already available to do this.
Best
Kami
Replies:
Posted on 11/07/14 07:50:27
Number of posts: 394
Dear PLS experts,
I ran PLS analysis with two regions (not a single voxel) as seeds. Each seed region included 27 voxels. I got two significant LVs, but I want to exclude the seeds (where the correlation between variables in X and Y are perfect) from the result. Could you please guide me to be able to exclude the seed regions? Since this could have been a common problem for many users, I hope a code is already available to do this.
Best
Kami
Hi Kami - perfectly correlated voxels are excluded from the analysis but its difficult to exclude all the voxels in a "seed region" unless you create a mask. A secondary issue is that even if you do that, the spatial autocorrelation (which is a combination of image processing artifact and real neurophysiology) is usually wider than the region size so there will be many voxels with high correlations
Practically speaking, whether the seed region is included in the analysis makes little difference to the overall statistical outcome (save for large bootstrap ratios), as the number of voxels outside the region is often much greater than those within. Also, it wil not affect the difference in functional connectivity between tasks, groups, or seeds.
Login to reply to this topic.
