ERP and behavioural PLS
giorgio.arcara
Posted on 03/08/12 07:09:14
Number of posts: 6
Dear PLS experts,
I wish to run a behavioural PLS on my ERP data.
I have data on says, 20 Subjects and 4 conditions. THe issue is that my behaviorual measure concerns Subject and so I have only 20 measures.
I tried to run a PLS analysis by replicating my behavioural measures 4 times, such as my behavioural file data fulfill the requirements.
In theory this is fine, because I'm interested to know if a single measure (on executive functions) correlates with scalp scores on the different conditions.
I'm wondering if this "replication" strategy is correct or if may lead to spurious results.
Thanks!
Giorgio
Untitled Post
I'm Online
jshen
Posted on 03/08/12 08:40:05
Number of posts: 291
Here's what we did in Behavioral PLS:
First, for each condition, we correlate data with their corresponding measures;
Then, we stack those correlated results to form a correlation map;
Finally, we use orthogonal transformation to find out the principal latent variable.
Usually, when we design behavioral experiments, we will record different measures that corresponding to different conditions, since the subjects are exposed with different stimuli.
Untitled Post
giorgio.arcara
Posted on 03/08/12 08:47:51
Number of posts: 6
I'm aware that usually there is a single behavioural measures for every single "Subject x condition" combination.
However I am interested to know if am allowed to do as I explained in my previous post, that is if it is possible (altough unusual) to use a behavioural measures as I explained.
Thanks
Giorgio
Untitled Post
I'm Online
nlobaugh
Posted on 03/08/12 09:03:47
Number of posts: 229
Hi Giorgio..
yes, this is something that you can do...
This kind of stacking is common when you have measures such as neuropsychological test scores, or some other measure not related to your particular experiment that you want to correlate with EEG/MRI/fMRI measures.
As long as your external measures are in the same row order as your subject data in the EEG datamat, you simply stack them for as many times as needed - in your case 4.
If you simply want to correlate the external measures with scalp scores from the results of a TaskPLS, you can do that as well, but not in the gui - you would need to extract the scalp scores from the result.mat file, and enter those into your favourite statistical package.
cheers,
Nancy