Is it true that if (X,Y) are bivariate normal then
|corr(X,Y)| .geq. |corr(a(X),b(Y))| for any functions
a and b? [where corr denotes the usual correlation
coefficient]
If so, can someone point me at a proof.
(I have a clunky almost-proof, but if the proposition
is true then I'm sure it will have been neatly proved
somewhere.)
If not, a counter example.
Thanks
Ray Watson
------------------------------------------------------------------
Ray Watson phone 03 8344 6411
Applied Statistics fax 03 8344 7499
Department of Mathematics & Statistics
University of Melbourne Vic 3010
Australia mailto:[log in to unmask]
------------------------------------------------------------------
|