Hi Donald and Helmut,
Thanks and let me describe my question and solution more clearly.
Initially, in 2 separate whole-brain simple linear regression models using variable V1 or V2 as the covariate, I found that V1 is correlated with activity in anterior portion of middle cingulate cortex (aMCC), and V2 correlated with its posterior part (pMCC). I want to do a direct statistical assessment of this difference in MCC.
I constructed another multiple linear regression model and entered V1 and V2 as the two covariates. Since I'm only interested in MCC, I performed small-volume correction (SVC) analysis in the contrast "1 -1", but the significance was borderline. When I used (i) contrast "1 0" and (ii) contrast "0 1", results were consistent with above analyses, i.e., V1 predicted aMCC activity after regressing out V2's effects, and V2 predicted pMCC activity after regressing out V1's effects. So my conclusion is "activity in different portions of MCC is predicted by V1 and V2."
I then extracted each individual's parameter estimates in the aMCC cluster (i.e., results from SVC) identified from (i), and plot 2 linear regression lines against V1 and V2 (this is outside spm). The correlation with V1 was significant, but not significant with V2. The slopes of the 2 lines were significantly different. The parameter estimates in the pMCC cluster identified from (ii) were also correlated with V1 and V2. The correlation with V2 was significant, but not significant with V1. However, the significance for the difference in slopes of these 2 lines was borderline.
I repeat Step 2 but, instead of using parameter estimates in the aMCC and pMCC clusters from SVC, I used data from clusters identified in the initial two simple linear regression analyses. Results here were the same as Step 2.
I wonder if Step 1 is sufficient (albeit not perfect because contrast "1 -1" is insignificant) to be a statistical assessment of difference in MCC. I'm also not sure if Steps 2 and 3 are appropriate, as well as the difference between Steps 2 and 3.
I do appreciate for help.