Hi,

I have a few questions about specifying a second level model.

I have a single group of subjects who went through one resting state scan. Each subject then performed a task under two conditions, so for each subject I have two behavioral measures score1 and score2, corresponding to the two conditions. 

1. I am interested in examining seed-to-voxel functional connectivity measures that are correlated with the change in the scores between the two task conditions. For example, I can hypothesize that a stronger fc with an ROI seed is correlated with a larger increase in score in Condition 1 (score1) compared to Condition 2 (score2).

I think I can specify 2nd level models in two ways.
First, I can specify the between subject effects [Allsubjects score1 score2 age] as [0 1 -1 0]. Alternatively, I can first calculate for each subject a difference score diff = score1 - score2, and then enter this score as a second-level covariate: [Allsubjects diff age] = [0 1 0]. I wonder which one of the specifications is appropriate?

2. Another related question: I also want to look at fc measures that are correlated with the score in general regardless of the condition (aka across conditions). In this case, if I have the second level as [Allsubjects score1 score2 age], is the correct specification [0 0.5 0.5 0]? Or an F-contrast [0 1 0 0; 0 0 1 0]? Or should I obtain the maps for each score separately ([0 1 0 0]) and [0 0 1 0] contrasts) and then use a conjunction in spm?

Thank you very much!

Hi,

I would just like to follow up to see if anyone can help me with these questions.

Thanks!