Within-subject, cross-trial regression

[@davidjustamante1449]

Thank you for trying to demonstrate how this computation is done. The Matrix representation starting at around 9:57 is a bit confusing confusing (and not correct). If you want to add the intercept it should be an Nx3 *3x1 + Nx1 = Nx1. The leftmost matrix should read (and assuming you want the intercepts to all be 1): [ e1_1 e1_2 e1_3 B1 1 y1 e2_2 e2_2 e2_3 [ B2 ] + [1] = [ y2] ... B3 1 ... en_1 en_2 en_3] yn

[@bokkieyeung504]

Hi Mike, in this video you plot the brain-behavior correlation in 2D and 3D ways. I would like to ask about the 2D plot: for each trial, we have power spectral information along different time points and frequencies, how did you produce a single power value for each trial?

[@mikexcohen1]

You could run the analysis for each pixel in the TF plot, or extract average power from within a TF window and use that for one regression analysis.

[@zmlin]

Hi Mike, I have my TF power and drawing speed (1 by time). Can I generate a similar correlation coeficient plot like in the video? How did you do it? I can only get one correlation coefficient per frequency using corr

[@mikexcohen1]

Yes, you would get a spectrum of correlations, not a TF plot. It would be frequency on the x-axis and correlation coefficient on the y-axis.