Non-rotated behvaioral PLS?
hsoderlund
Posted on 02/19/08 13:37:27
Number of posts: 0
Hi there,
I was wondering if it is possible to run a non-rotated behavioral PLS, so that for example you can pre-specify that all your behaviors are related to a brain pattern in the same way, in order to identify brain areas involved in all these behaviors.
Cheers,
Hedvig.
Untitled Post
I'm Online
nlobaugh
Posted on 02/19/08 14:15:55
Number of posts: 229
Hedvig..
If what you are asking could be rephrased as: Is there a brain pattern that equally supports all behaviour measures of interest? BehavPLS actually answers that question for you.
It may be the case that all behaviours have positive (negative) correlations with brain activity (most likely LV1). One example of this is in our ERP paper (Neuroimage, 2005, 26:986, Fig 5). In this case, the behaviour was reaction time, so the common correlation pattern is perhaps not surprising. The remaining LVs describe the differences in brain-behaviour relations.
In the situation where the input behaviours are not measuring a single 'domain' (or correlated domains) you are less likely to find a common pattern across behaviours.
nancy
Untitled Post
hsoderlund
Posted on 02/19/08 14:23:38
Number of posts: 0
Thanks Jimmy and Nancy,
The question could indeed be rephrased as you said, Nancy. The behavioral PLS that I ran show an LV1 where all behaviors but one go in the same direction, and no (significant) LV show them all in the same direction. That's why I was wondering if I could "force" them all into the same direction. But as you say, if they're not measuring similar features it will be less likely to find overlapping brain patterns.
Maybe I should assume, then, that there is no brain patterns common to all included behaviors.
Thanks,
Hedvig.