Hi there,

We (the Jagust Lab) would like to clarify the interpretation of the results of a simple behavioral PLS. I thought I had this down but now after discussions with my colleague Beth Kuczynski I'm not so sure any more. Let's assume we have a set of resting FDG PET scans (1 scan per patient) for a group of patients and we collected a bunch of behavioral measures (MMSE, age etc.). Now we run a behavioral PLS with one measure at a time, let's say MMSE. (Randy had confirmed that such an approach is valid).

Now we look at our results and we have a significant and reliable LV. Further we get a positive correlation of MMSE with the LV. First question: how do we know the exact significance level of plotted correlation [brain scores for behavioral variable]? Further, if we have a positive correlation, only the yellow regions (positive salience) in the bootstrap ratio plot represent a positive relationship between brain scores and the behavioral variable (positive correlation) . However, only the blue regions (negative salience) represent a negative correlation. So in other words, yellow (positive) saliences are showing an increasing covariance pattern with increasing MMSE or decreasing covariance with decreasing MMSE. Moreover blue (negative) saliences will show an increase in covariance with a decrease of MMSE or a decrease of covariance with increasing MMSE. Finally, to confirm, if in a separate analysis we happen to get a negative correlation the yellow (positive) salience then represent a negative correlation between brain score and behavioral variable while the blue now represents a positive relationship.

Many THANKS & greetings from California!

Ansgar & Beth