Using 2ndlevel design_matrix for confounds and regressors of interest

[JohannesWiesner]

Might be related to my post.