Hello PLS experts,

I'm new to PLS and am using the command line version of the software to analyze event-related fields. These are not sensor-level data, they are the time courses from vertices in a dSPM map but, for the sake of generalizability, these vertices could be considered analogous to sensors, and I will use that term below. I am performing a repeated measures comparison of 19 subjects and, for simplicity, let's say I'm only comparing two experimental conditions. Each ERF consists of 376 samples at 20482 vertices/sensors. I am conducting a mean-centered task PLS analysis using all default options for mean centering ... not that I think that matters for this particular question.

When I create my data matrix to enter into PLS, it is formatted as such:

- rows = conditions x subjects (first all subjects in condition 1, then all subjects in condition 2)

- columns = samples x sensors (first all sensors at the first sample, then all sensors at the second sample, etc.)

In the case of my analysis, this is an 38 x 7701232 matrix. I know that the brain saliences are stored in result.u. In my case, result.u is a 7701232 x 2 matrix because there are 2 LVs. Is it safe to assume that the 7701232 rows in my result.u matrix are organized in the exact same way as the columns were in my original data matrix? Specifically, do the rows represent samples x sensors? I have tried to determine this by inspecting the PLS code but it is a bit beyond my comprehension.

Many thanks,

Lauren