0
$\begingroup$

If $X_1,X_2$ constitute a random sample of size n=2 from a Poisson Population show that the mean of the sample is a sufficient estimator of the parameter $\lambda$ .

Since the sum of Poissons is also Poisson with parameter $\lambda+\lambda=2\lambda$
Then $E[Y]=2\lambda$. where Y constitute of two Poisson distributions each with parameter $\lambda$.

Do I have to show that $2\lambda$ is sufficient for $\lambda$.
I don't understand what the estimator is here?

$\endgroup$
1
$\begingroup$

Probably you mean that the sample average is sufficient statistics for $\lambda$. So, if $X_1, X_2$ are independent draws from Poisson distribution with parameter $\lambda$ then the answer is "yes".

Check here: http://en.wikipedia.org/wiki/Sufficient_statistic


Using Fisher–Neyman factorization theorem:

$P\{X=x\}=P\{X_1=x_1,X_2=x_2\}=P\{X_1=x_1\}P\{X_2=x_2\}=\lambda^{x_1}e^{-\lambda}/x_1!*\lambda^{x_2}e^{-\lambda}/x_2!=\lambda^{(x_1+x_2)/2}e^{2-2\lambda}/(x_1!x_2!)=g_\lambda(T(x))h(x)$

where $h(x)=(x_1!x_2!)^{-1}$ and $g_\lambda(T(x))=\lambda^{(x_1+x_2)/2}e^{2-2\lambda}$

$\endgroup$
2
$\begingroup$

You're not showing sufficiency at all. The only thing your calculation shows is that $\bar X = (X_1 + X_2)/2$ is an unbiased estimator of $\lambda$.

To show that $\bar X$ is a sufficient statistic, you have to do something else. The most straightforward approach is to use the Factorization Theorem, which seeks the expression of the joint distribution for $X_1, X_2$ as $$f_{X_1, X_2}(x_1, x_2) = g(T(x_1, x_2) \mid \lambda) h(x_1, x_2),$$ for some function $T$ that depends only on the sample; and in your case, $T(x_1, x_2) = \bar x = (x_1 + x_2)/2$.

$\endgroup$
  • $\begingroup$ I didn't knew that the estimator that I should show that is sufficient is $(x_1+x_2)/2$. $\endgroup$ – clarkson Apr 9 '14 at 5:41

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.