Unless you have many outcome variables that you need to analyse together, you don't need MANOVA, you just need to use the REPEATED statement in a normal GLM. This takes account of the fact that your observations are correlated because you rated the same test subject 3 times. There is a good overview at the attached link.

Do you have lots of outcome measures?

http://www.ats.ucla.edu/stat/sas/seminars/sas_repeatedmeasures/

Thanks Clare. I only have one outcome. 

A simple SAS code you may want to start with:

PROC GLM data=Your Dataset;

   CLASS   group;

   MODEL Y1 Y2 Y3 = group;

    MANOVA H=group ;

RUN;

A website I liked most:

http://psych.colorado.edu/~carey/courses/psyc7291/examplecode.htm

Thank you so much Guiping. It is exactly what i was looking for