Let $X_1, X_2, ..., X_n$ be a random sample from $Binom(1, p)$. I'm trying to find the UMVUE of $p^3$.

Some thoughts:

  1. Apparently, $\bar{X}^3$ is not the answer, although it's the MLE of $p^3$.
  2. For distinct $i$, $j$, and $k$, the distribution of $X_iX_jX_k$ is $Binom(1, p^3)$, but what if $n$ is not divisible by three?
  • $\begingroup$ What is an unbiased estimator of $p^3$? $\endgroup$ – StubbornAtom Apr 20 '19 at 15:48
  • 1
    $\begingroup$ See math.stackexchange.com/q/2687375. And if the answers there are helpful you might answer this question yourself. $\endgroup$ – StubbornAtom Apr 20 '19 at 16:14
  • $\begingroup$ @StubbornAtom Thanks for the information; they are really helpful! Wasn't aware of the Lehmann–Scheffé theorem. +1 $\endgroup$ – nalzok Apr 20 '19 at 16:32
  • $\begingroup$ Please provide an answer to your question so that it goes off the unanswered list. $\endgroup$ – StubbornAtom Apr 23 '19 at 10:33
  • 1
    $\begingroup$ @StubbornAtom Done! $\endgroup$ – nalzok Apr 23 '19 at 12:16

Actually, this problem is a classic example of the Lehmann–Scheffé theorem. The theorem states

If a statistic that is unbiased, complete and sufficient for some parameter $\theta$, then it is the UMVUE for $\theta$.

Here $\theta$ is $p^3$, and $T = \sum_{i=1}^n X_i$ is a sufficient and complete statistics for $p^3$, so we simply need to construct a unbiased estimator of $p^3$ with $T$. In other words, we need to find $\phi$ such that $E(\phi(T)) = p^3$. For example, you can readily verify

$$ \phi(T) = \frac{T(T-1)(T-2)}{n(n-1)(n-2)} $$

And that's the UMVUE of $p^3$.

| cite | improve this answer | |

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.