I've had a number of requests concerning what answers I got to my problem
of generating a set of orthogonal comparisons.
Well, I got two sets of answers before going to bed last night and both
worked. Nobody really gave me an algorithm for doing them - I guess it's a
seat-of-the-pants thing. The two things I learned were to break it down
into subsets where the n-1 restriction applies, and not to try to use
fractional coefficients, which boggle the brain too quickly when there are
a lot of them. I'm a little relieved that some real statisticians also
consider the problem non-trivial, because my self-esteem was rather
battered last night.
Thanks to anyone who has replied since; I haven't checked all my mail yet
this morning. Integer coefficients - that's the ticket.
david
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|