Dear SPM experts
I have the question about the issue of unbalance sample size in two group covariate interaction analysis. The following statement was what i did in the voxel wised statistical model.
I have two study groups (Group 1: healthy control N=30 and Group 2: patient group N=50). I want to investigate the age * group linear interaction effect in the voxel wised manner (I want to investigate the group difference in the slope of age-volume correlation between these two study groups). I constructed the GLM model as follows (using 3 NC and 5 Patient as examples ):
EV1(NC group) EV2(Patient group) EV3 (NC-age-demean) EV4 (Patient-age-demean)
1 0 -16.25 0
1 0 -6.25 0
1 0 3.75 0
0 1 0 -16.25
0 1 0 -6.25
0 1 0 3.75
0 1 0 13.75
0 1 0 23.75
The contrast of this GLM model were as follows:
Positive interaction (0 0 1 -1) and negative interaction (0 0 -1 1)
Due to the different sample size of each groups, does the statistical results of positive and negative interaction are biased to this statistical design ?
Best
Paul
|