Hi Hui Xu, You can do the 2-way mixed ANOVA in randomise. Do as described in the manual with the subtractions. There are alternative set-ups that will lead to the same results and which have been discussed many times in the mailing list. Consider doing a search in the archives. See for example this one: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;3cee56eb.1510 All the best, Anderson On 14 June 2017 at 04:17, Hui Xu <[log in to unmask]> wrote: > > Dear Expert: > > I have 2 groups data at time 1 and time 2. I've read this model: > *http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29* > <http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29> > > But the last messages: > Due to how the data would need to be permuted, the FEAT model may not be > used in randomise. Instead, just as in the paired t-test example, paired > differences within-subject would be computed via fslmaths and a two-sample > t-test > <https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#Two-Group_Difference_.28Two-Sample_Unpaired_T-Test.29> > could be used to test whether the run1-run2 difference differed between the > two groups. > > So I can't do 2-way Mixed Effect ANOVA in randomise, what can I do to > analysis the 2 group* 2 timepoint data? > And I set up the model, how can I use it in the later analysis? > > Thanks! > Hui Xu >