Suppose that $(x_1,..., x_n)$ is a sample from a Poisson($λ$) distribution with $λ ≥ 0$ unknown. If we use the prior distribution for $\lambda$ given by the $Gamma(a,b)$ distribution, then determine the posterior distribution of $\lambda$.

I believe I need $\pi(\lambda | s) = L(\lambda | \bar{x}) \times \pi(\lambda)$

likelihood function times prior

likelihood of poisson is given by

$L(\lambda | \bar{x}) = \prod_{x=1}^{n} \frac{\lambda^x e^{-\lambda}}{x!} = \frac{1}{(nx!)} \lambda^{\sum_{x=1}^{n} x} e^{-n\lambda} = \frac{1}{(nx!)}\lambda^{n \bar{x}}e^{-n\lambda}$

prior is given by

$\pi (\lambda)$ ~ $Gamma(a, b)$

$= \frac{b^a \lambda^{a-1} e^{-b\lambda}}{\Gamma (a)}$ which is the probability density function

so finally

$\pi(\lambda | s) = L(\lambda | \bar{x}) \times \pi(\lambda)$

not sure how to compute this.

Any help appreciated !


The posterior density is given by $$ f_{\lambda\mid x}(\lambda\mid x)\propto L(\lambda)\times f_{\lambda}(\lambda) $$ where $f_{\lambda}$ is the prior for $\lambda$ and $L$ is the likelihood. In our case $$ L(\lambda)\times f_{\lambda}(\lambda)\propto\lambda^{\sum_{i=1}^n x_i}e^{-n\lambda}\times \lambda^{a-1}e^{-b\lambda}=\lambda^{a-1+\sum_{i=1}^n x_i}e^{-\lambda(n+b)} $$ where we dropped all the constants (including those that depend solely on $x$). Hence $$ f_{\lambda\mid x}(\lambda\mid x)=C\lambda^{a-1+\sum_{i=1}^n x_i}e^{-\lambda(n+b)} $$ where $C$ is a function of $x$. By inspection $\lambda\mid x$ is a Gamma distributed random variable i.e. $\lambda\mid x \sim \text{Gamma}(a+\sum_{i=1}^n x_i, n+b)$ as desired.


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.