Thanks for the reply. However, I think I am accounting for the repeated measures in my design. The original design matrix I sent was an attempt to combine the 1x4 repeated measures model and the 2x2 ANOVA model on the feat webpage. EVs 4-11 should remove the subject means and then EV1 models main effect 1, EV2 main effect 2 and EV3 the interaction. The design matrix looks almost identical to the 2x2 ANOVA example with the exception of EV4, which in the example appears to be modeling the overall intercept for the group - the model I am using models each subject mean instead (EVs 4-11). The other difference between the models is the inclusion of the f-contrast matrix - including the --fc=design.fts flag results in the following error:
An exception has been thrown
Runtime error:- detected by Newmat: matrix is singular
MatrixType = Crout # Rows = 1; # Cols = 1
Trace: Crout(lubksb); GeneralSolvI; InvertedMatrix::Evaluate; Gsmanager::ols_contrasts; Gsmanager::ols; Gsmanager::run.
When I exclude the f-contrast matrix, flameo runs fine, except that it uses 21 degrees of freedom.
Thanks again,
Eric
|