On Dec 11, 1:20 pm, paul <paulh...@infi.net> wrote: > Does a multiple regression with all dummy (indicator) variables make > sense?
Thanks for the replies! What puzzles me is that the book used for the course (Keller, but I don't have a copy handy for a more detailed reference) does devote considerable space to methods for use with or instead of ANOVA to make pair-wise comparisons. Since the book gives details of how to use Excel for computations, I'm surprised the all dummy variable technique isn't mentioned. It is certain;y easy to implement. Does the technique conform to the assumptions that the residuals are normally distributed with equal variance?