Conditional Expected Value Problem involving Poisson and Gamma distributions

288 Views Asked by At

The question is

An insurance company supposes that the number of accidents that each of its policyholders will have this year is Poisson distributed, with a mean depending on the policyholder: the Poisson mean $\Lambda$ of a randomly chosen person has a Gamma distribution with the $\Gamma(2, 1)$-density function $f_\Lambda(\lambda) = \lambda{e^{-\lambda}} $ ($\lambda > 0$). Find the expected value of $\Lambda$ for a policyholder having $x$ accidents this year ($x = 0, 1, 2, \ldots$)

I am very unsure of how to solve this problem. I know that we need to solve for $E(\Lambda\mid X=x).$ I am pretty sure, but not sure what comes after this.

1

There are 1 best solutions below

4
On

For any fixed $x=1,2,3,\ldots$ $$ \mathbb E(\Lambda\mid X=x) = \int_0^\infty \lambda\cdot f_{\Lambda\mid X}(\lambda\mid x)\, d\lambda $$ is the expected value of a random variable with conditional pdf $$ f_{\Lambda\mid X}(\lambda\mid x) \propto f_{\Lambda}(\lambda)\cdot f_{X\mid \Lambda}(x\mid \lambda). $$ Here the second term is the pmf since $X$ is discrete r.v. $$ f_{X\mid \Lambda}(x\mid \lambda) = \mathbb P(X=x \mid \Lambda=\lambda) = \frac{\lambda^x}{x!}e^{-\lambda} $$ and therefore $$ f_{\Lambda\mid X}(\lambda\mid x) \propto f_{\Lambda}(\lambda)\cdot f_{X\mid \Lambda}(x\mid \lambda)=\lambda e^{-\lambda} \cdot \frac{\lambda^x}{x!}e^{-\lambda}\propto \color{red}{\lambda^{x+1} e^{-2\lambda}}. $$ So conditional distribution of $\Lambda$ given $X=x$ is $\Gamma(2,x+2)$. Indeed, the pdf of $\Gamma(2,x+2)$ is $$ \gamma_{2,x+2}(\lambda) = c\cdot \color{red}{\lambda^{x+2-1}e^{-2\lambda}} $$ This is exactly the pdf that we got.

The expected value this distributon is $\frac{x+2}{2}$. So the answer is $$ \mathbb E(\Lambda\mid X=x) = \frac{x+2}{2}. $$