Back to PLS HelpPosted on 09/17/14 14:30:23
Number of posts: 5
Hi all,
I am working on a pretty complex dataset and am just wondering if anyone knows whether or not a PLS analysis would be appropriate / could work. The data consists of 2 conditions (rest vs. active), 2 participant groups (control vs feedback), 2 sessions ( 1 vs. 2), 14 participants total (7 per group), 15 frequency bands, and 3160 possible Functional Connections (based on a coherence analysis).
I am just wondering if there is anyway to use a PLS to test differences in Functional connectivity between conditions and groups, across sessions.
I have done a PLS before using a 2 (conditions) x 19 (participants) x 15 (frequency bins) x 3160 neural connections. That one was more straight forward. However I have a feeling that throwing in the groups and sessions makes a PLS not usable. If anyone has any insight or suggestions on how to analyze this data that would be great.
Thanks a lot!
-Ron
Replies:
Posted on 09/17/14 16:53:47
Number of posts: 394
Hi all,
I am working on a pretty complex dataset and am just wondering if anyone knows whether or not a PLS analysis would be appropriate / could work. The data consists of 2 conditions (rest vs. active), 2 participant groups (control vs feedback), 2 sessions ( 1 vs. 2), 14 participants total (7 per group), 15 frequency bands, and 3160 possible Functional Connections (based on a coherence analysis).
I am just wondering if there is anyway to use a PLS to test differences in Functional connectivity between conditions and groups, across sessions.
I have done a PLS before using a 2 (conditions) x 19 (participants) x 15 (frequency bins) x 3160 neural connections. That one was more straight forward. However I have a feeling that throwing in the groups and sessions makes a PLS not usable. If anyone has any insight or suggestions on how to analyze this data that would be great.
Thanks a lot!
-Ron
Hi Ron - you can certainly use PLS for this, but you may have to run a few analyses to get an idea of what you have. I might suggest something like:
1) run the whole thing to get a feel for the number of stable effects
2) if #1 yields good results, break it down along a reasonable dimension (e.g., group) to see what comes out. You can think of this as a sort of post-hoc test
3) consider putting in apriori contrasts of interest to evaluate specific effects (non-rotated PLS)
I'd be interested to hear how this goes, so please let me know!
cheers
Randy
Login to reply to this topic.
