2
$\begingroup$

The following Corollary is used in "Theory of Point Estimation" by Lehmann to prove a theorem. However I'm unsure how to prove this Corollary (it's left as a problem, so proof is omitted). The corollary states:

A necessary and sufficient condition for a statistic $U(X)$ to be sufficient is that for any fixed $\theta_1, \theta_2$ , the ratio $p_{\theta_1} (x)/ p_{\theta_2}(x)$ is a function only of $U(X)$.

Clearly if $U(X)$ is sufficient, then the ratio is a function only of $U(X)$ as a direct result of the factorisation theorem.

However I'm not sure how to prove that $p_{\theta_1} (x)/ p_{\theta_2}(x)$ being a function only of $U(X)$ implies sufficiency. Intuitively we would expect this to be the case, also because of the factorisation theorem, but I'm not sure how I'd prove this formally.

$\endgroup$
1
$\begingroup$

Let $R$ be the support of the distribution of $X$, and suppose that do not depend on the parameter $\theta$. Fix one value of $\theta_2$, and let $\theta_1$ vary over the parameter space. Write $p_{\theta_2}(x)= g(x)$ (for clarity). Then we get $$ p_{\theta_1}(x)=g(x) h(U(x);\theta_1) $$ for some function $h$, and $g$ (not depending on $\theta_1$). But that is the factorization theorem.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.