 |
 |
Dear Kathleen, if your model fits the data sufficiently well, you should not need to change the default initial F-threshold. There is nothing really wrong with raising the threshold, but you should definitely look at your data and model to check whether something might have gone wrong in your data (outlier subjects) or model specification (e.g. images shuffled between subjects, conditions or time points, unexpected estimates of error covariance etc). If you included multiple images per subject (e.g. one per time point), it might be worth trying to compute a single image per subject (first level fMRI or ImCalc) to simplify error variances in your anova model. Hope this helps, Volkmar