Hello PLS wizards,
I recently conducted a study that compared brain activity between 2 states (rest and task), across 19 participants, in 60 frequency bins, with 3160 possible neural connections (2 x 19 x 60 x 3160 dataset). The results of my PLS show 2 latent variables; 1) There is a difference between states and 2) There is no difference between states. For both LVs, I get a significant result, p<.002 for 1, and p<.01 for 2.
My (possibly incorrect) interpretation of the results are that indeed there are neural connections that are significantly different during Task then during Rest ( i.e. a specific task-positive network), and that there are ALSO neural connections that are the same between both states. I go on to explore the functional connectivity in ONLY the Task minus Rest condition to highlight the connectivity during task only.
Is this logic correct? Is it ok to focus on one variable only as the purpose behind my project is to look only at the task-positive network.
Any help/advice would be greatly appreciated.
Thanks! - Ron
hi Ron,
How did you run the analysis? I assume this was done from the command line rather than the GUI? Also, what mean-centering option did you use? If you use the default, then the second LV in your analysis can be ignored b/c technically it does not exist. You can check this by looking at its singular value (s). If its zero, then you can ignore it. The reason its 'significant' has to do with the precision in MATLAB, where there can be a non-zero at the gazillionth decimal place. We probably should put a mask function where is a singular value is zero, its not considered.
If you did not use mean centering then indeed you have one LV that id's differences in functional connectivity and one that id's similarities. You can focus on the differences, if that is your only question.
hi Ron,
How did you run the analysis? I assume this was done from the command line rather than the GUI? Also, what mean-centering option did you use? If you use the default, then the second LV in your analysis can be ignored b/c technically it does not exist. You can check this by looking at its singular value (s). If its zero, then you can ignore it. The reason its 'significant' has to do with the precision in MATLAB, where there can be a non-zero at the gazillionth decimal place. We probably should put a mask function where is a singular value is zero, its not considered.
If you did not use mean centering then indeed you have one LV that id's differences in functional connectivity and one that id's similarities. You can focus on the differences, if that is your only question.
Hey Randy,
Thanks for the reply. I used the MATLAB script, not the GUI. I chose a mean-centred, then the other option where I set the permutation and BSR to 2^9.
So based on your reply I guess that means I have 2 significant LVs, of which I can focus on the differences. Correct me if I'm wrong.
Thanks again!
-Ron
Actually, I think you have just one, but to be sure what are the values in "result.s"? If the second is zero, then you have one significant LV. This is because mean-centering removes the grand mean, which would identify the similarities between tasks.
Hey Randy,
For results.s, the1st number is 11.3, and the 2nd is 3.4 e-4.
-Ron
Hey Randy,
For results.s, the1st number is 11.3, and the 2nd is 3.4 e-4.
-Ron
ok - so you have 1 LV. The second should have a singular value of 0, but as I said, MATLAB doesn't like absolute zero
Thanks a lot for your help Randy!
I will keep this in mind during future studies.
-Ron
Baycrest is an academic health sciences centre fully affiliated with the University of Toronto
Privacy Statement - Disclaimer - © 1989-2024 BAYCREST HEALTH SCIENCE. ALL RIGHTS RESERVED