Expected value and Variance of an MLE

390 Views Asked by At

I have the poisson distribution $X_a\sim $Poisson$(a\mu)$ for $a=1....n$ and I have found the maximum likelihood estimator as:

$$\hat{\mu}=\frac{\sum_{a=1}^{n}{x_a}}{\sum_{a=1}^{n}a}$$

However I am not sure what the best way to find the expected value of this is?

I would assume that $E(\hat{\mu})=\frac{x}{a}$ and $Var(\hat{\mu})=\frac{a\mu}{n}$ but I am not too sure. I was wondering if anyone can check this and see if I have gone wrong?

1

There are 1 best solutions below

0
On BEST ANSWER

Assuming independence...

The likelihood is the following

$$L(\mu) \propto e^{-\mu\frac{n(n+1)}{2}}\mu^{\Sigma X}$$

the loglikelihood is

$$l(\mu)=-\mu\frac{n(n+1)}{2}+\Sigma X log\mu$$

Derivating...

$$l^*(\mu)=-\frac{n(n+1)}{2}+\frac{\Sigma X}{\mu}$$

Setting the score=0 and solving it in $\mu$ you get

$$\hat{\mu}_{ML}=\frac{2\Sigma X}{n(n+1)}$$

observing that $\Sigma X \sim Po\Big(\frac{n(n+1)}{2}\mu\Big)$ you get

$$E(\hat{\mu})=\mu$$

...it's unbiased.

I let you calculate its variance (it is very very easy)...

EDIT:

$$V(\hat{\mu})=\frac{2\mu}{n(n+1)}$$