2
$\begingroup$

I'm studying the Rayleigh distribution given by $$ f_v(x) = \frac{x}{v} \ \exp(-\frac{x^2}{2v}) \ \mathbb{1}\{x \geq 0\}$$

and found an estimator (using method of moments) to be $$ \tilde{v} = \frac{2}{\pi n^2} \left(\sum_{i=1}^{n} X_i \right)^2$$

Now I want to know whether it's biased or not, but I get stuck when I try to compute the expectation of $\tilde{v}$. I have $$ \mathbb{E}[\tilde{v}] = \frac{2}{\pi n^2} \mathbb{E}\left[\left(\sum_{i=1}^{n} X_i \right)^2\right]$$

I tried stating that $\mathbb{E}[X] = \frac{1}{n}\sum_{i=1}^{n} X_i$, which yields $\left(\sum_{i=1}^{n} X_i \right)^2 = n^2 \mathbb{E}[X]^2$ and an unbiased estimator.

However, that seems wrong, as that statement is exactly what I used to get the estimator in the first place, so it seems like I'm reasoning in circles here. Is it wrong? If so, how else could I get the bias?

Thanks in advance!

$\endgroup$
  • $\begingroup$ Yes, the exercise insisted on using $v$ instead of $\sigma^2$, which complicated matters somewhat. However, taking the square root of my estimator results in the estimator given in that link, so I assumed it was correct (as $v = \sigma^2$ in this parametrization), . $\endgroup$ – Tiamo P. Oct 5 '16 at 12:21
  • $\begingroup$ Revision of earlier comment, to which your responded: To start, please see http://math.stackexchange.com/questions/738519/method-of-moments-and-maximum-likelihood-question and the Wikipedia article on 'Rayleigh distribution'. You are using parameter vv while the usual parameterization is in terms of $v = \sigma^2$. $\endgroup$ – BruceET Oct 7 '16 at 16:23
3
$\begingroup$

Recall that $var(X)=\mathbb{E}X^2-\mathbb{E}^2X$, so \begin{align} \mathbb{E}\tilde{v} &= \frac{2}{\pi n^2} \mathbb{E}\left(\sum_{i=1}^{n} X_i \right)^2 = \frac{2}{\pi n^2}var(\sum_{i=1}^{n} X_i )+\frac{2}{\pi n^2}\mathbb{E}^2(\sum_{i=1}^{n} X_i)\\ &= \frac{2}{\pi n^2}n\sigma_X^2+\frac{2}{\pi n^2}(n\mu_X)^2\\ &= \frac{4-\pi}{\pi n}v+ v \neq v,\,\,\, \forall n\in\mathbb{N} \end{align}

$\endgroup$

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.