Thank you Anderson! I have some minor follow-up questions:
1. Running my model using the setup I suggested did not produce a rank-deficiency error. Why is this?
2. I would like to also obtain mean maps of activation for each group separately (patients and controls). Is this possible in the model you have put forward?
3. Can you also clarify (given the instructions on http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM) why it is not necessary to demean the categorical variables in my model?
4. lastly, I would like to explore diagnosis by gender interactions. In this case, would the following setup be appropriate?
EVs:
EV1: male healthy (0 or 1)
EV2: female healthy (0 or 1)
EV3: male patient (0 or 1)
EV4: female patient (0 or 1)
EV5: age (continuous, demeaned)
EV6: scan site 1 (0 or 1)
EV6: scan site 2 (0 or 1)
EV6: scan site 3 (0 or 1)
EV6: scan site 4 (0 or 1)
EV6: scan site 5 (0 or 1)
contrasts:
main effect of diagnosis: [1 1 -1 -1]
main effect of gender: [1 -1 1 -1]
diagnosis by gender: [1 -1 -1 1]
Thank you again!
Lara
|