Unbiasedness of tests for homogeneity (Q580840)

The authors consider the problem of testing homogeneity of k-samples drawn from a one-parameter exponential family with pdf \(f(x,\theta)=\beta (\theta)e^{\theta x} h(x)\) where h(x) is assumed log concave or is a Polya frequency function of order two. W.l.g. we can assume k samples of size one each and \(T=\sum^{k}_{i=1}x_ i\) is complete sufficient under \(H_ 0.\) The authors consider unbiased permutation invariant tests and these are conditional on \(T=t\) in view of the Neyman structure property. The authors show that any size \(\alpha\) permutation invariant test, whose conditional (on \(T=t)\) acceptance sections are convex, is unbiased. They also prove that any size \(\alpha\) test which is Schur-convex for fixed t is unbiased.