I have this problem that asks to prove that an estimator is biased:

Here's the problem:

Let $X_1, X_2, ... X_n$ be a random sample from an exponential distribution with Poisson rate $\lambda > 0$ (and mean $1/\lambda$). Consider the estimator $\hat \Lambda = \frac{1}{\bar X} = \frac{n}{\sum X_i}.$

I'm supposed to show that it's biased, find MSE and find a $C$ that will make it unbiased but this is what I got:

$E(\hat\Lambda) = E(\frac{n}{\sum X_i})$

  • 2
    $\begingroup$ It seems that you think that $\mathsf E\frac1{Y}$ equals $\frac1{\mathsf EY}$. That however is not true in general. $\endgroup$ – drhab Nov 22 '17 at 16:15
  • $\begingroup$ Can you please explain why? Doesn't it fall under the property $E(aX) = aE(X)$? $\endgroup$ – iamnobody Nov 22 '17 at 16:25
  • $\begingroup$ No, it does not fall under linearity of expectation. Note that $x\mapsto\frac1{x}$ is not a linear map. It will explain itself if you try it out on some random variable $Y$ and in a certain sense you are asked to do that ("prove that the estimator is biased"). $\endgroup$ – drhab Nov 22 '17 at 16:31
  • $\begingroup$ Thank you for pointing that out! That will be a really big help for me. However, I'm still stuck solving this problem. I'm guessing I have to integrate it instead to get the expected value? But I've never encountered integrating a summation before so I'm unsure how to proceed. $\endgroup$ – iamnobody Nov 22 '17 at 17:11

Here is one way to do it. First, start by noting that a sum of iid exponential random variables is a gamma random variable. More precisely, if $X_1, \ldots, X_n \sim \exp(\lambda)$, then $$\sum X_i \sim \textrm{Gamma}(n, 1/\lambda)$$ Setting $Y := \sum_{i=1}^n X_i$ for a moment, this means that the PDF of $Y$ is: $$ f_Y(y) = \frac{1}{\Gamma(n)(1/\lambda)^n} y^{n-1} e^{-y/(1/\lambda)} $$ Then the expected value of $1/Y$ is the following integral: $$ E(1/Y) = \int_0^{\infty} \frac{1}{\Gamma(n)(1/\lambda)^n} y^{n-1} e^{-y/(1/\lambda)} \cdot \frac{1}{y} \; dy$$ Doing a little algebra here and using the fact that $\Gamma(n) = (n-1)!$, you can rewrite this as $$ E(1/Y) = \frac{\lambda}{n-1} \cdot \int_0^{\infty} \frac{1}{\Gamma(n-1)(1/\lambda)^{n-1}} y^{(n-1)-1} e^{-y/(1/\lambda)} dy$$ Now the trick is to recognize that the integrand is just the PDF of a Gamma($n-1$, $1/\lambda$) random variable, so it integrates to one and therefore you're just left with $E(1/Y) = \lambda/(n-1)$. Now you can do the rest.


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.