Hi Yawu,

Probably the safest approach is to do compute t2-t1 for each subject who has both timepoints, and also t3-t1 for each subject, and do a simple 2-group comparison (as in a two-sample, unpaired t-test) using either MANOVA or NPC, the latter having the advantage of being more powerful and allowing for directional effects. In PALM you'd run as this:

palm -i diff_t2-t1.nii -i diff_t3-t1.nii -d design.mat -t design.con -npc -n 5000 -approx tail -nouncorrected -o myresults -logp

It is also possible to do a 2x3 repeated measures ANOVA, assuming compound symmetry between all three timepoints. In that case, please see a design here: https://s3-us-west-2.amazonaws.com/andersonwinkler/mailinglist/design_yawu.ods

All timepoints would be used. PALM would be invoked as this:

palm -i all_timepoints.nii -d design.mat -t design.con -f design.fts -n 5000 -approx tail -nouncorrected -o myresults -logp

Hope this helps!

All the best,

Anderson

On 26 December 2017 at 03:34, Liu Y <[log in to unmask]> wrote:

Hi Anderson,

We are interested in the difference among the three time points in each group, and also the difference in DTI change between two groups.

Best regards,

Yawu

2017年12月26日 06:35，"Anderson M. Winkler" <[log in to unmask]>写道：

Hi Yawu,

Are you interested in differences between timepoints and whether these differ between groups, or are you interested on the average between timepoints and whether these differ between groups? Or are you interested in something else?

All the best,

Anderson

On 21 December 2017 at 03:21, Liu Y <[log in to unmask]> wrote:

Dear Experts,

We have 2 groups, three time-points: how to model all time-points within a single design given that not all subjects have second or third time-point measurements, and take into account baseline differences in DTI metrics.

Any suggestions are highly appreciated.

Best regards,

Yawu