 |
 |
Dear SPM experts, I have a question regarding the plot option in SPM visualization panel for displaying the results of a VBM study. In a full factorial model with 1 factor in 2 level , 3 covariates of no interest (adjusting the data for them - age, total GM volume and total WM volume) and one covariate (X) with interaction with our factor (and centered on the overall mean). if I plot the fitted response for the interaction of the covariate and group (differences in slope of the regression line in groups), and I plot the adjusted values versus scan, then the values in the variable y (in the workspace) are the adjusted gray mater volume in a specific voxel corrected for (age, total GM volume and total WM volume) and not corrected for covariate X and so if I regress it for each group with the corresponding X values ([ones(length(X1)), X1]) I should end up with the b values (regression values) equal to the beta values of the SPM for that contrast and for that voxel (which comes out of the spm_graph.m)? Then I should be able to show the interaction effect with the slope differences of the changes of gray mater volume in that specific voxel with the changes in covariate X in 2 groups. what is wrong with the above description? As the regression slopes that I get by regressing the adjusted values (y) with the values of covariate X in Matlab, are not equal to the beta values from SPM and even do not have the same sign! And if "adjusted response" here means adjusted for the effect of other covariates but not the X covariate? Many thanks in advance, Shahrzad