Hello!
I was playing around with different EVs in an event related design and thinking about my design matrix, and I noticed that all of my EVs were scaled to have the same peak amplitude, even if some were repeated in succession more/more often than other. In other words, the peaks of my isolated events in one EV would be scaled to be the same height as the peak of repeated events in a separate EV. I looked inside the design.mat and found the same thing in the numbers. Is this a visualization technique, but not actually reflected within the glm calculations themselves? (That is my current bet). Or is design.mat the actual matrix that is acted upon?
The reason I ask is because it seems like scaling all EVs to have the same peak would hurt one's ability to estimate the betas for EVs whose predicted timecourse did not naturally reach that peak; by scaling it up to match other EVs, the ultimate beta will be reduced by that of the scaling factor. Your wisdom on this would be greatly appreciated!
|