Hi Anderson,
I am trying to set up a 2 (group, between sub) X 3(emotion, within sub) model, where I am interested in the group x emotion interaction.
I am following advice from Mumford Brain Stats fb group to set up this model (https://www.facebook.com/groups/mumfordbrainstats/), but if this is the not correct set up then I would greatly appreciate your input on how to run this.
Here are the directions that I received, copied here:
----
"Basically, scroll down a little to the "Factor Effects" version of the 2x3 ANOVA here
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM...
Obviously, this model isn't what you'd want to directly use, but you can tweak it to adjust for the repeated measures. The 3 level variable is your within-subject covariate, while the 2 level represents your two groups. The first step is to remove the "mean", column of 1's (last regressor) and replace it with a mean for each subject. In other words, each subject will have their own column that is all 0's, but 1's for the 3 valence measures. This addresses repeated measures, but now you have a rank deficient design. Specifically, using the subject-specific regressors, you can recreate the "A" regressor. The solution is to get rid of the A regressor. You will not be able to use this model to test for the main effect of "group" To do that, you'd need a different model. The primary purpose of this model is to test your interaction.
Last, you can specify the F-tests, F2 and F3, as they have done. F3 is the interaction effect. If the interaction is not significant, then remove it from the model and make your life easier. If it is significant, then I'm willing to bet what you've already done helps explain the interaction.
ANOVA: 2-factors 2- & 3-levels (2x3 between-subjects ANOVA) - GLM - FslWiki
FSL.FMRIB.OX.AC.UK"
---
Once I have run this model successfully using flameo, I am having trouble with using the "smoothest" and "cluster" commands, as stated in the previous message. Another related question, my DLH smoothness estimate from "smoothest" is 0.04 (typed incorrectly in last message). Is 0.04 an appropriate / realistic value? I am switching from AFNI, where the output of 3dFWHMx represented the smoothness in mm (e.g 6mm-10mm).
Smoothest also requires DOF. I used df = 140: 219 copes (3 conditions x 73 subjects) - 77 regressors (4 between sub, and 73 representing the subject specific regressors for the mixed effects model). Is this the correct way to calculate the DOF for this mixed effects analysis?
Thanks,
Michelle
|