Repeated Measures meets Latin Squares Crossover Design Repeated Measures meets Latin Squares
Layout for Crossover design Group I II Subject 1 … 9 10 18 Time A B 2 Layout for Crossover design
First nine subjects at Time 1 Group Time Drug Y 1 A 41.9 2 35.1 3 38.6 4 36.1 5 34.6 6 39.7 7 37.8 8 38.8 9 39.1 First nine subjects at Time 1
Crossover model (split–plot univariate analysis) Yijkl=µ+Gi+S(i)j Between Subjects +Tk+Dl+εijkl Within Subjects Crossover model (split–plot univariate analysis)
Enter all terms in the Model as Fixed to get all SS from JMP
This gives all Model and error SS
If we specify Subject as Random
Subject(Group) tests Group, Residual tests the Within terms
Usually hope Time is not significant, but at least we controlled for it
What would happen if we believed our Model and Tested Drug A at Time 1 and Drug B at time 2? Hint is next slide…..
Consider the Drug effect
Plot Time*Drug
Look at A at time 1 and B at time 2, they are about equal.
Moral: Crossing is good Main Effects would have been confounded Moral: Crossing is good
Actually only four distinct Predicted values. Why?
Now check Normality