Dear Prof. Zalesky

When i tried to model the group term in a repeat measure design as well as the subject mean, NBS will always give me a 'rank deficient' warning. Does that mean I have to get rid of the group term if I am not interested in the group effect? Or if I am interested in the group effect, I don't need to model the subject mean?

Thanks a lot!

John

Hi John,

Group terms should be removed. There is no need to include the group effect in a group by time design where time is the repeated measure. A repeated measures design is not needed to test for a group effect.

Regarding your earlier question, the design matrix looks ok, although it seems you have included two columns for the the linear effect. You only need one of the those columns. The polarity (+1 or -1) of the contract vector will dictate whether the linear effect is a positive or negative correlation. Remember to use exchange blocks for a repeated measures design.

Andrew


Originally posted by john zang:

Dear Andrew

Thank you for your quick response! Its really helpful

John