Back to PLS HelpPosted on 10/27/14 17:14:18
Number of posts: 2
Dear PLS experts,
I am writing to ask whether in behavioral PLS there are any restirctions on what type of behavioral data can be put in. I know the PLS deals with multicolinearity between the behavioral variables, however, what if the behavioral variables are orthogonal to each other? I am using behavioral performance scores that were created with PCA to yield summary scores for some broader cognitive constructs. I wonder if there is any caveat in this. The result is a single LV in which two out of four of these orthogonal components show the same direction of associations with the brain data. I will be greateful for your thoughts on it.
With best regards,
Aga
Replies:
Posted on 10/27/14 18:09:57
Number of posts: 394
Hi Aga - there are no restrictions on the nature of the behavior data that you put in. In your case, even if you use scores from a PCA, this does not mean the LVs from PLS will align with each of the components independently. It is quite possible for two or more PCA components to contribute to the same LV - the advantage you have is that you can make the case of the unique contribtion of each component to the prediction of the brain data captured on the other side of the LV.
Hope that makes sense
Randy
Posted on 10/27/14 18:13:11
Number of posts: 2
Hi Randy,
Thank you very much for the clarification, it does make sense!
best,
Aga
Login to reply to this topic.
