Dear Chih-Hao, To model the group by variable X interaction, you need to select "Interactions" "With Factor 1" when specifying the covariate X in the interface (and Centering option "Factor 1 mean") and use contrast [0 0 1 -1]. See also: http://mumford.fmripower.org/mean_centering/ Best regards, Guillaume. On 22/12/16 04:31, 連 志浩 wrote: > Dear experts, > > > We want to investigate interaction between Groups (ex. normal and > patients) and a continuous variable X. > > I think I should do these followed steps: > > > 1. Use 2nd-level modeling, and choose "Two-sample t-test". > > 2. Input normal subjects' .con image into "Group 1 scans" and patients' > .con image into "Group 2 scans", then set Variance as "Unequal". > > 3. Create a covairate (variable X) for all subjects. > > 4. Run 2nd-level modeling and estimate its' SPM.mat file. > > 5. Reulst -> difine a t-contrast -> vector [ 0 0 1] > > > then I can get the interaction between Group and Variable X. > > > Is anything wrong is my steps? > > or there's some better ways to analysis the interaction I want to > investigate? > > > Thanks for your reading. > > > Best, > > Chih-Hao > > -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG