Finding UMVUE, given complete and sufficient statistic

303 Views Asked by At

Let $X_1,X_2,…,X_n(n≥3) $ be a random sample from Poisson(), where$ θ∈(0,∞)$ is unknown and let

$T=\sum _{i=1}^n X_i$

The UMVUE of $ e^{−2\theta}*\theta^3$ is

Since UMVMUE is a Complete Sufficient Statistics then

$E(ϕ(t))=e^{−2\theta}*\theta^3$

$\sum \phi(t)∗ \frac{e^{−nθ}*(nθ)^t}{t!}=e^{−2\theta}*\theta^3$

$\sum \phi(t)∗ \frac{(nθ)^t}{t!}=e^{−2\theta+n\theta}*\theta^3$

$\sum \phi(t)∗ \frac{(nθ)^t}{t!}=\sum \frac{(−2\theta+n\theta)^r}{r!}*\theta^3$

How do we proceed after this?

1

There are 1 best solutions below

0
On

First, to use a standard notation, let's set

$$S=\sum_i X_i$$

and observe that $S$ is CSS, Complete and Sufficient Statistic.

Then let's construct an unbiased estimator for $g(\theta)=e^{-2\theta}\theta^3$

A very simple unbiased estimator is the following:

$$T=\mathbb{1}_{(X_1\cdot X_2=2)}$$

where it indicates that $T=1$ if $X_1\cdot X_2=2$ and it is zero elsewhere.

Immediately you can verify that $\mathbb{E}[T]=e^{-2\theta}\theta^3$

Now you can apply Rao-Blackwell theorem calculating

$$\mathbb{E}[T|S]$$

which is your UMVUE, given that $S$ is also complete.


$$\mathbb{E}[T|S]=\mathbb{P}[T=1|S=s]=\frac{e^{-2\theta}\theta^3\mathbb{P}[S=s|T=1]}{\mathbb{P}[S=s]}=$$

$$=\frac{e^{-2\theta}\theta^3\mathbb{P}\left[ \sum_{i=3}^n X_i=s-3\right]}{\frac{e^{-n\theta}(n\theta)^s}{s!}}=\frac{e^{-2\theta}\theta^3\cdot \frac{e^{-(n-2)\theta}[(n-2)\theta]^{s-3}}{(s-3)!}}{\frac{e^{-n\theta}(n\theta)^s}{s!}}=\frac{(n-2)^{s-3}}{n^s}\cdot\frac{s!}{(s-3)!}$$