The answer to this depends on whether your covariates change in time or not. If they change through the session then you need to set something up in the 1st level design. If they apply to the whole session then you should set something up in the 2nd
level design. The latter is more common and easier, but the former can be done as well, although you won't find as many examples as the exact implementation depends on the style/timing of the stimulation.
Dear FSL experts,
In my experiment, the regressor of interest contains 4 conditions (i.e., the combination of 2 stimulus types and 2 task types) (i.e., 4 EVs). My objective is to compare activations between the 2 stimulus types and between the 2 task types, and to analyze
their interaction. However, behavioral data show that two rating scores (i.e., ratings for the level of distractability and level of confidence), which are not of interest but possibly confound activations, are significantly different among these 4 conditions.
For each of these 4 conditions, each subject has 1 averaged score for the level of distractability and 1 averaged score for the level of confidence.
Thus, I want to put these scores as covariates in my analysis, i.e., to eliminate their potential confounding effects on fMRI data. Could someone tell me how? Should I put these scores as covariates in 1st-level or 2-level FEAT?
Thank you. Mike