Dear FSl experts,
i want to perform a 2X2 repeated measures anova (2 display conditions x 2 reward levels), but there is not exact example in FSL ANOVA designs. Can you please describe it in details, how to perform it? I am especially interested in interaction between factors.
I have tried 4 approaches,
1. I have created desing matrix for participants like this example:
subject1 subject2 main effect 1 main effect 2 interaction
cope 1 subject 1 1 0 1 1 1
cope 2 subject 1 1 0 1 -1 -1
cope 3 subject 1 1 0 -1 1 -1
cope 4 subject 1 1 0 -1 -1 1
cope 1 subject 2 0 1 1 1 1
cope 2 subject 2 0 1 1 -1 -1
cope 3 subject 2 0 1 -1 1 -1
cope 4 subject 2 0 1 -1 -1 1
Running analyses with mixed effect and I obtained results with 27 dof as stated in the stats/dof file.
2.
I tried also another but similar approach like this:
subject1 subject2 main effect 1 subject 1 main effect 2 subject 1 Interaction subject 1 main effect 1 subject 2 main effect 2 subject 2 Interaction subject 2
cope 1 subject 1 1 0 1 1 1 0 0 0
cope 2 subject 1 1 0 1 -1 -1 0 0 0
cope 3 subject 1 1 0 -1 1 -1 0 0 0
cope 4 subject 1 1 0 -1 -1 1 0 0 0
cope 1 subject 2 0 1 0 0 0 1 1 1
cope 2 subject 2 0 1 0 0 0 1 -1 -1
cope 3 subject 2 0 1 0 0 0 -1 1 -1
cope 4 subject 2 0 1 0 0 0 -1 -1 1
But the analyses with mixed effect did not run, aborte with an error (An exception has been thrown
nevs >= ntpts, Singular matrix.Trace: Gsmanager::flame_stage1; Gsmanager::run.)
The analyses run only with fixed effect, and I obtained zero dof as stated in the stats/dof file.
3. In the last approach, i have treated copes separately, in particular, we created two main contrasts, for first contrast, we contrasted regressor 1 with regressor 2, for second contrast, we contrasted regressor 3 with regressor 4. As a next step, we contrasted these two newly created contrasts.
I obtained 9 dof.
4. In the last approach i treated subjects separately, creating separately gfeat for each like this:
subject1 main effect 1 main effect 2 interaction
cope1 1 1 1 1
cope2 1 1 -1 -1
cope3 1 -1 1 -1
cope4 1 -1 -1 1
And then i took resulting copes and i treated them for all subjects together, but separately for main effect1, main effect 2 and interaction, and I obtained 9 dof.
What is the correct way to perform it?
And another related question- how can i figure out the direction of interaction from the results?
I have additional question regarding no- interest regressor related to accounting for RT differences in different task conditions. Should i account for RTs differences in the first level analysis, putting it as and additional no interest regressor, with 3 column- format- second column as a RT, or third column as a de-meaned RT? Or another option, should i model RTs differences in higher level steps of analyses?
Thanks for a feedback.
Best,
Jana
|