Hi Kim,
Presumably you are testing for differences in those 8 cases,
so it is necessary to model each one separately. And adding
a collapsed regressor to this model (collapsed across your 8
conditions) would not be an option, as it would be collinear
with the other 8 regressors (it would equal their sum).
So if your goal is to compute the average response, the glt
method should be a good way to do it.
- rick
p.s. I'm glad you enjoyed the class. :)