Dear Karl, Donald, and other F-contrasts wizards,

is it possible to test 1st level F-contrasts for group effects at the 2nd level? And if yes, what test would I have to use? What is the assumed null distribution for 1st level ess-images? Or would I have to use randomization and/or permutation techniques to build an empirical null?

Brief background: I have 3 parametric modulators derived from a computational model that are linearly dependent (they are all probabilities that add up to 1) and that I have entered them unorthogonalized into the design matrix. I expect that they will explain variance in certain brain regions and I have computed appropriately weighted F-contrasts spanning the 3 regressors. Now I want to test for group effects at the 2nd level ...

Thanks for your insights.

Jan