Sorry but I think more information might be necessary to clarify the question in my last post. Please forget it and see the attached file.
In brief, I set up PPI model in each subject to see which brain regions showed different connectivity with a seed region (called ROI) under different experimental context ("on" vs "off") (Fig. 1). I took the 1st-level "PPI on - PPI off" (higher connectivity during "on" condition than during "off" condition) con file of each subject (Fig. 2) to a group analysis, where a simple linear regression analysis was performed (Fig. 3). Each subject has one behavior measure (called "VRS"), and we wanted to identify interregional functional connectivity that is positively correlated with "VRS" (Fig. 4). (I think my PPI procedures should be correct)
The whole-brain analysis yielded a significant cluster in insula (I binarized this mask and saved as "insula-cluster"), meaning that ROI-insula coupling was correlated with "VRS." What I want to do is extract the ROI-insula connectivity and plot a simple linear regression figure, with "VRS" as x axis and ROI-insula connectivity as y axis. Below is my methods and I hope you experts can check my procedures:
(1). I added a new F contrast to account for the effect of covariate "VRS" (see Figure 4-2).
(2). loaded the 2nd-level contrast by "Results" (uncorrected p value = 1, which selected the whole brain).
(3). press "eigenvariate" then adjusted data for the F contrast (i.e. the effect of covariate).
(4). select the "insula-cluster".
Then I got 17 numbers and the simple linear regression plot showed beautiful correlation with "VRS". My questions are:
(a). Are my procedures correct?
(b). In Step (1), should the F contrast be [1 0;0 1] to account for the effect of both group mean and covariate?
(c). In (3), if I did not adjust data for the F contrast, there would be no correlation, which contradicted whole-brain analysis results.
(d). Are the 17 numbers "adjusted parameter estimates" or "parameter estimates" (I believe it's the former)? Can I just call it "ROI-insula connectivity" in my simple linear regression plot?
Thank you so much in advance.
Mark
|