0
$\begingroup$

We have $n$ independent observations, $\{Y_1,...,Y_n\}\overset{iid}\sim\mathcal{Poissson}(\mu)$. Consider the estimator $\tilde\mu = \bar Y$. I have to show that $\tilde\mu$ is unbiased and find the variance. Did I do this correctly?

$$\begin{align}\mathbb{Bias}(\tilde\mu) &= \mathbb E\tilde\mu-\mu\\ &= \mathbb E(\tfrac 1n\sum_{i=1}^n\mathbb E(Y_i))-\mu\\ &= \tfrac 1n\sum_{i=1}^n\mathbb E(Y_i)-\mu\\&=\mu-\mu\\&=0 \\ \\\therefore\quad \mathbb{Var}(\mu) &=\mu&&\text{Since }\{Y_1,...,Y_n\}\overset{iid}\sim\mathcal{Poissson}(\mu)\end{align}$$

$\endgroup$
2
$\begingroup$

The variance is incorrect. Note that $$ \text{Var}(\bar{Y})=\frac{1}{n^2}\sum \text{Var}(Y_i)=\frac{1}{n^2}\times n\mu=\frac{\mu}{n} $$ since the $Y_i$ are independent.

$\endgroup$
  • $\begingroup$ Thank you! Also, if I were asked if Mu(~) is a consistent estimator of Mu, how would I tackle this? $\endgroup$ – Chance Gordon Feb 3 '18 at 1:28
  • $\begingroup$ @ChanceGordon You've shown that $E\bar{Y}=\mu$ when you showed that the bias is zero. Hence $\bar{Y}$ is a consistent estimator of $\mu$. $\endgroup$ – Foobaz John Feb 3 '18 at 1:35
  • $\begingroup$ Make sure you can distinguish the terms consistent estimator and unbiased estimator. For unbiasedness, John has already argued in the above comment. For consistency, you need to show that $\bar{Y} \to \mu$ in probability, which is easy after you calculating the variance as you can now use the Chebyshev's inequality and the variance term will squeeze the probability to $0$. $\endgroup$ – BGM Feb 3 '18 at 2:46

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.