Back to PLS HelpPosted on 02/16/16 10:34:19
Number of posts: 14
Hello,
I have performed a mean centering PLS analyses with 5000 permutations and 500 bootstrapping samples to detect the structural differences between two groups of subjects. And i observe that LV1 is significant with P<0.05 and shows that there is a pattern corresponding to the group difference in the brain structure. But when i use a bootstrapping threshold of 2.57, i do not see any reliable clusters that survive this threshold.
How do i interpret this result?
1) Even though LV1 significantly shows that there is a group difference, should i still conclude that there is no reliable spatial location corresponding to the pattern shown by LV1? Is it that the pattern is a global effect in the brain and cannot be pinned to a specific spatial location?
or
2) Does it mean that the pattern shown by LV1 does not exist any where in the brain?
Best Wishes
Jay
Replies:
Posted on 02/16/16 10:38:40
Number of posts: 394
Hello,
I have performed a mean centering PLS analyses with 5000 permutations and 500 bootstrapping samples to detect the structural differences between two groups of subjects. And i observe that LV1 is significant with P<0.05 and shows that there is a pattern corresponding to the group difference in the brain structure. But when i use a bootstrapping threshold of 2.57, i do not see any reliable clusters that survive this threshold.
How do i interpret this result?
1) Even though LV1 significantly shows that there is a group difference, should i still conclude that there is no reliable spatial location corresponding to the pattern shown by LV1? Is it that the pattern is a global effect in the brain and cannot be pinned to a specific spatial location?
or
2) Does it mean that the pattern shown by LV1 does not exist any where in the brain?
Best Wishes
Jay
Hi Jay - there is nothing magical about the threshold 2.57. You can use 2.3 or even 1.96 so long as you are clear on what you are infering from your results. Its entirely possible to get a broad effect that at an individual voxel level seems weak, but is sufficiently relable to produce a multivariate effect
Login to reply to this topic.
