Dear experts,
We would like to have some insights on ways to test our research hypotheses. Our experimental design has one categorical variable (group) with three levels and one continuous physiological variable. We would like to test possible group difference across regression coefficients between a physiological measure and brain activity under a certain task condition.
H0= b1 = b2 = b3
At the second level we thought of running a multiple regression with 5 predictors:
1 = dummy variable for group A (all 1 for subjects in group A and rest 0)
2 = dummy variable for group B (all 1 for subjects in group B and rest 0)
3 = physiological * dummy_A (physiological measure * dummy_A variable) = b1–b2–b3
4 = physiological * dummy_B (physiological measure * dummy_B variable) = b2 –b1–b3
5= physiological measures = b3
Is the following F contrast adequately test the main effect ?
F contrast 0 0 1 1 0 or 0 0 ½ ½ 0
To decompose the effect our idea was to run three separate t tests.
Our interrogations
1)We would like to know if our thinking is right with respect to the predictor and contrast.
2)If you have better ideas on proper t contrast that could decompose the effect within this model.
Thanks in advance,
Annie
|