Hello fellow CONN people,
I was wondering what the scale may denote when one uses the "Average Connectivity w/ Cluster Contrasted Across Subjects" graphic function but the subject factor isn't a group-wise contrast but a correlation/regression?
For instance, I ran a MVPA regression (0 1 0 0) and found a
significant cluster. When I select the graphic function
option using this cluster as a seed and the same regression model
weights, the scale is -0.06 to 0.06 r and the legend states it
reflects a diff r, but again the subject-wise model used to
generate the maps was correlation w/ my values-of-interest (i.e., 0
1 0 0; AllSubj, Values-of-interest, Age, QAfactor).
Thanks for any help or thoughts,
Jeff Browndyke
Hi Jeff,
Those plots always show the effect-size of your GLM model, so for a regression analysis (e.g. with factors [AllSubects, SeverityScore, Age, QAfactor] and a [0,1,0,0]), the color will scale with the values of regressor coefficients associated with the second factor in that model and estimated at each voxel. The units of those coefficients in a regression model are in "difference in r per unit change in your regressor" (e.g. r-change divided by SeverityScore-change), so you are right the "r diff" name is a bit misleading, sorry.
Hope this helps
Alfonso
Originally posted by Jeff Browndyke:
Hello fellow CONN people,
I was wondering what the scale may denote when one uses the "Average Connectivity w/ Cluster Contrasted Across Subjects" graphic function but the subject factor isn't a group-wise contrast but a correlation/regression?
For instance, I ran a MVPA regression (0 1 0 0) and found a significant cluster. When I select the graphic function option using this cluster as a seed and the same regression model weights, the scale is -0.06 to 0.06 r and the legend states it reflects a diff r, but again the subject-wise model used to generate the maps was correlation w/ my values-of-interest (i.e., 0 1 0 0; AllSubj, Values-of-interest, Age, QAfactor).
Thanks for any help or thoughts,
Jeff Browndyke