2
$\begingroup$

Consider iid random varibales $(X_i)_{1\le i\le n}$ with $X_i\sim Exp(\theta)$ for $1\le i\le n$ and $\theta\in(0,\infty)$. Then we have $$\sum_{i=1}^nX_i\sim \Gamma(n,\theta)$$ with density function

$$f(x)=\frac{\theta^n}{(n-1)!}x^{n-1}e^{-\theta x}\mathbb{1}_{[0,\infty)}(x).$$

Now I want to calculate

\begin{align}E\Bigg[\frac{1}{\frac{1}{n}\sum_{i=1}^nX_i}\Bigg]=n\cdot E\Bigg[\frac{1}{\sum_{i=1}^nX_i}\Bigg]&=n\cdot\int_{0}^\infty \frac{1}{x}\frac{\theta^n}{(n-1)!}x^{n-1}e^{-\theta x} dx\\ &=\frac{n\theta^n}{(n-1)!}\cdot\int_{0}^\infty x^{n-2}e^{-\theta x} dx,\end{align}

but I do not know where I have to go from here.

I know the antiderivate of $x^{n-2}e^{-x}$, but I do not know how to deal with $x^{n-2}e^{-\theta x}$. Can someone give me a hint?

Thanks in advance!

$\endgroup$
  • $\begingroup$ Change of variables? $\theta x = y$. $\endgroup$ – d.k.o. Nov 5 '18 at 23:33
  • $\begingroup$ Yes, that's what I was looking for. Thank you! $\endgroup$ – user408858 Nov 5 '18 at 23:54
1
$\begingroup$

Use the fact that $$\int_0^\infty x^m e^{-ax}\,dx=\frac{\Gamma(m+1)}{a^{m+1}}$$ In your case, you would get \begin{align}\frac{n\theta^n}{(n-1)!}\int_0^\infty x^{n-2} e^{-\theta x}\,dx& =\frac{n\theta^n}{(n-1)!}\frac{\Gamma(n-1)}{\theta^{n-1}} \\ &= \frac{n\theta}{(n-1)!}(n-2)! \\ &= \frac{n\theta}{n-1} \end{align}

$\endgroup$
1
$\begingroup$

You don't need to go into calculus for this one. We will use the Gamma pdf instead. I'll continue from where you stopped: $$\frac{n\theta^n}{(n-1)!}\cdot\int_{0}^\infty x^{n-2}e^{-\theta x} dx=\frac{n\theta}{(n-1)}\frac{\theta^{n-1}}{((n-1)-1)!}\cdot\int_{0}^\infty x^{(n-1)-1}e^{-\theta x} dx=\frac{n\theta}{(n-1)}$$ Since the $\frac{\theta^{n-1}}{((n-1)-1)!}\cdot\int_{0}^\infty x^{(n-1)-1}e^{-\theta x} dx$ part is also a Gamma pdf, just with the parameter $n-1$ instead of $n$, basically specifically $\Gamma (n-1,\theta)$, it reduces to just $1$

$\endgroup$
1
$\begingroup$

Set $Y=X_1+\dots+X_n$.

You can do it by integration by parts (and induction). Here are two other methods.

  1. Use the identity $\frac{1}{y} = \int_0^\infty e^{-s y} ds$ to obtain \begin{align*} E [ 1/Y] &= E [\int_0^\infty e^{-s Y} ds] \\ & = \int_0^\infty E[e^{-s Y}] ds \\ & = \int_0^\infty E[e^{-s X_1}]^n ds\\ & =\int_0^\infty (\frac{\theta}{\theta+s})^n ds \\ & = \theta^n \int_\theta^\infty u^{-n} du \\ & = \frac{\theta}{n-1} \end{align*}

  2. Differentiation under integral sign. \begin{align*} E[1/Y] &= \frac{\theta^n}{(n-1)!} \int_0^\infty \frac{1}{y} y^{n-1} e^{-\theta y} dy\\ & = \frac{\theta^n}{(n-1)!} (-1)^{n-2}\frac{d^{n-2}}{d \theta^{n-2}}\int_0^\infty e^{-\theta y} dy\\ & = \frac{\theta^n}{(n-1)!} (n-2)! \theta^{-(n-1)}\\ & = \frac{\theta}{n-1} \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.