6.29 If a minimal sufficient statistic exists, a necessary condition for a sufficient statistic to be complete is for it to be minimal. [Hint: Suppose that T = h(U) is minimal sufficient and U is complete. To show that U is equivalent to T , note that otherwise there exists

ψ such that ψ(U) = η[h(U)] with positive probability where η(t) = E[ψ(U)|t].]