Dear Vincent,
"it separates the block from a linear increase with mean 0"
Not sure whether I understand this correctly, but the global regressor accounts for the average activation in both the simple and the PM model, the PM additionally accounts for the linear change (+/- relative to the average) in the PM model. This is the common way to test for linear effects of PMs. In case you want to look at the activation level for the minimum value of the PM (or any other value) you could combine the betas, global beta - x * PM beta.
Best,
Helmut
|