Dear SPM user/experts,
I'm new to SPM and have a question about 2nd level analysis and covariates. I looked into the list and the manual but handling and entering covariates is only coarsely described.
I have a repeated measures design with only one group and multiple conditions. Let us assume I conducted an experiment with two conditions (A and B). I collected data from 10 participants. As covariate, I have 3 different behavioral rating variables (C1-C3) collected for conditions A and B.
First I want to know whether activation in condition A covaries with any of the covariates. One analysis would be to run a one-sample t-test. I use the scans from condition A and then I enter e.g. only covariate C1. After entering the Vector and name of the covariate I can specify the interactions.
Question 1: What are the factors 1, 2, & 3 I can choose for one-sample t-test covariates? Is "factor 1" my condition A? Should I choose the option “none” for a one-sample t-test? Why do results change when choosing interaction “none” compared to choosing “with factor 1”.
Question 2: What would happen if I enter all covariates in this step at once (C1+C2+C3 instead of only C1)? I heard that the estimations for the second and third covariate are only based on part of the data, i.e. the part which variance has not been explained by the first covariate (case for C2) and the part which variance has not been explained by the 1st and 2nd covariate (case for C3). Is this correct?
Next I want to use a paired t-test and compare conditions A and B and how they interact with covariate C1 (meaning I would have 2 covariate vectors: a vector A_C1 and B_C1).
Question 3: Would I now choose different a different factor for each covariate vector in the “interactions” option? If so, would the scans of condition A be equal to “factor 1” and the scans of condition B be equal to “factor 2”?
As far as I know these questions are related to all 2nd level analyses be it regression models or t-tests.
Thanks a lot for your help!
Felix
|