Back to PLS HelpPosted on 05/04/17 12:46:54
Number of posts: 2
I ran a regular behavior PLS to investigate the relation between brain (ERP data) and behavior. In the scalp scores plot I see that there is a correlation of .73 and of .85 for my two conditions. But when I look at the behavioral LV both conditions show a behav LV of about .6 and the p-value is not significant. What does this behavioral LV tell me? And how do I know whether my brain-behavior correlation is a true effect?
Thank you very much in advance,
Kalinka.
Replies:
Posted on 05/04/17 12:57:23
Number of posts: 394
I ran a regular behavior PLS to investigate the relation between brain (ERP data) and behavior. In the scalp scores plot I see that there is a correlation of .73 and of .85 for my two conditions. But when I look at the behavioral LV both conditions show a behav LV of about .6 and the p-value is not significant. What does this behavioral LV tell me? And how do I know whether my brain-behavior correlation is a true effect?
Thank you very much in advance,
Kalinka.
the behavlv is the singular vector from the SVD scaled by its corresponding singular value. It will be similar to the pattern of correlations of each behavior with the scalp/brain scores, but not identical. If the entire LV is non-significant by permutation (the p-value) then its not 'true" effect in the sense of being differentiable from a random pairing of the EEG data to behavior. The correlations themselves do have confidence intervals calculated but if the overall relation is not significant, it makes the interpretation of the correlations and confidence intervals difficult
Posted on 05/05/17 07:02:30
Number of posts: 2
the behavlv is the singular vector from the SVD scaled by its corresponding singular value. It will be similar to the pattern of correlations of each behavior with the scalp/brain scores, but not identical. If the entire LV is non-significant by permutation (the p-value) then its not 'true" effect in the sense of being differentiable from a random pairing of the EEG data to behavior. The correlations themselves do have confidence intervals calculated but if the overall relation is not significant, it makes the interpretation of the correlations and confidence intervals difficult
Thank you very much for the clear and quick answer!
Kalinka
Login to reply to this topic.
