Back to PLS HelpPosted on 06/27/17 11:41:16
Number of posts: 13
Hi there,
I'm wondering how to extract the brain score value associated with each condition in my LV from the results.mat file. I understand that brain scores are stored in the result.usc file, but I'm not sure how these values map onto the brain scores obtained from 'Taks PLS Brain Scores With CI' plot. Any help would be much appreciated.
Best wishes,
Heather
Replies:
Posted on 06/27/17 13:49:05
Number of posts: 394
Hi Heather
The brainscores are stored as subjects witthin condition within group. The plots are condition/group averages of these scores with bootstrap estimated confidence intervals. These may be mean-centred depending on the analysis you did, but the relative pattern of the mean will be preserved.
Posted on 06/28/17 05:51:02
Number of posts: 13
Hi Heather
The brainscores are stored as subjects witthin condition within group. The plots are condition/group averages of these scores with bootstrap estimated confidence intervals. These may be mean-centred depending on the analysis you did, but the relative pattern of the mean will be preserved.
Hi Randy,
Thanks for the speedy reply. After having a look at the average brain scores in the results.usc file I'm not actually finding the same pattern seen on the plots. Specifically the fourth condition is assigned a negative brain score whose confidence interval does not cross the x-axis on the plots but a positive average according to the individual brain scores, while the sixth condition is assigned a positive brain score whose confidence interval does not cross the x-axis on the plots but a negative average according to the individual brain scores. We did indeed use the mean-centred approach, but subtracting the mean from the individual brain scores still results in a pattern inconsistent with the plots. Do you have any more insights into what my be going on?
Best,
Heather
Posted on 06/28/17 10:55:12
Number of posts: 394
Hi Randy,
Thanks for the speedy reply. After having a look at the average brain scores in the results.usc file I'm not actually finding the same pattern seen on the plots. Specifically the fourth condition is assigned a negative brain score whose confidence interval does not cross the x-axis on the plots but a positive average according to the individual brain scores, while the sixth condition is assigned a positive brain score whose confidence interval does not cross the x-axis on the plots but a negative average according to the individual brain scores. We did indeed use the mean-centred approach, but subtracting the mean from the individual brain scores still results in a pattern inconsistent with the plots. Do you have any more insights into what my be going on?
Best,
Heather
thanks heather - I will need to look at your results - can you send me the results file? rmcintosh@research.baycrest.org
Posted on 06/29/17 10:59:11
Number of posts: 394
thanks heather - I will need to look at your results - can you send me the results file? rmcintosh@research.baycrest.org
Hi Heather - I took a look at the results and the scores seem fine. The means that are plotted are results.boot_result.orig_usc which is derived from results.boot_result.usc2. The averages here show a simiar relative pattern as results.usc (the difference has to do the mean centering).
Does that make sense or am I missing something?
Posted on 06/29/17 11:26:55
Number of posts: 13
Hi Heather - I took a look at the results and the scores seem fine. The means that are plotted are results.boot_result.orig_usc which is derived from results.boot_result.usc2. The averages here show a simiar relative pattern as results.usc (the difference has to do the mean centering).
Does that make sense or am I missing something?
Hi Randy - definitely makes sense and thank you for taking a look. While the pattern of results is similar, the average of my fifth condition is still coming up as negative according to the results.boot.orig_usc but positive in the plots. This definitley influences how I interpret my results. Do you know what may be underlying this difference or which I should rely more heavily on? I'm also asking because I was hoping to re-create the plots generated by PLS myself using the exact LV data for a paper figure.
Best,
Heather
Posted on 06/29/17 11:38:29
Number of posts: 394
Hi Randy - definitely makes sense and thank you for taking a look. While the pattern of results is similar, the average of my fifth condition is still coming up as negative according to the results.boot.orig_usc but positive in the plots. This definitley influences how I interpret my results. Do you know what may be underlying this difference or which I should rely more heavily on? I'm also asking because I was hoping to re-create the plots generated by PLS myself using the exact LV data for a paper figure.
Best,
Heather
odd - if you calculate the means they all show the same relative pattern (see below)
Posted on 07/04/17 11:17:00
Number of posts: 13
odd - if you calculate the means they all show the same relative pattern (see below)
Ah I see now! thank you for posting that. I didn't realize the means were stored only in the first column. Definitely have the info I need to create beautiful figures now. What are the other 5 columns for if you don't mind my asking? I assumed they corresponded to the different lags of my task.
Best,
Heather
Posted on 07/04/17 11:20:42
Number of posts: 394
Each column is a latent variable (LV). The last one will be degenerate since you mean-centred the data, hence the singular value is zero.
Posted on 07/04/17 11:35:25
Number of posts: 13
Each column is a latent variable (LV). The last one will be degenerate since you mean-centred the data, hence the singular value is zero.
Beauty! Thank you again for all the help. Love using your software.
Best,
Heather
Posted on 07/04/17 11:52:24
Number of posts: 394
Thanks! :)
Login to reply to this topic.
