Thank you very much, yes that cleared it up. I actually didn't see that 2 nuisance columns (as only one group is affected) would render the design rank deficient - so that's cool as it doesn't leave me with a choice :) Regarding the OLS, what I meant, and assumed was actually running either of the 2 models through OLS would be the same as the variance is modeled differently (compared to FLAME).
Anyway, thanks for your patience, Torsten
|