Hi Anderson,
The design matrix contains a regressor for each combination of factor 1 and 2 e.g., EV1= factor1 value=1, factor2 value=1; EV2=1,2; EV3=2,1; EV4=2,2; EV5=3,1; EV6=3;2. Factor1 has 3 levels, factor2 has 2 levels. I want to test whether the factor2 effect varies parametrically with factor1 i.e., whether (EV2-EV1) > (EV4-EV3) > (EV6-EV5).
My original question was how to set this contrast up at the first level.
After I posted the question, I found the following useful link: http://mumford.bol.ucla.edu/lin_cont_illus.pdf. At the third level, I could set this up as a [1 0 -1] linear contrast of the EV2-EV1, EV4-EV3, EV6-EV5 contrasts from the second level.
Does that sound right?
|