3
$\begingroup$

Suppose $Y_1, \dots Y_n$ are exponentially distributed: $Y_i | \lambda \sim Exp(\lambda)$. Find the conjugate prior for $\lambda$, and the corresponding posterior distribution. Show that the posterior mean for the failure rate $\lambda$ can be written as a weighted average of the prior mean and the maximum likelihood estimator, $\hat{\lambda}=\bar{y}^{-1}$.

Given that the joint pdf as follow:-

$P(Y_i | \lambda)=\lambda^{n}e^{-\lambda \sum_{i=1}^{n} Y_i}$

Then the likelihood function is

$ L (\lambda | Y_i)\propto \lambda^{n}e^{-\lambda \sum_{i=1}^{n}Y_i}$

which give Gamma distribution with $\alpha = n+1$ and $\beta =\sum_{i=1}^{n} Y_i$.

Thus, with the prior density

$P(\lambda)\propto \lambda^{a-1} e^{- \lambda b}$

gives the posterior distribution as follow:-

$P(\lambda | Y_i) \propto \lambda^{n+a-1} e^{-\lambda (\sum_{i=1}^{n}Y_i + b)}$

and resulting in Gamma distribution with $\alpha = n+a$ and $\beta = \sum_{i=1}^{n} Y_i + b$.

My problem now is how can I show the posterior mean for the failure rate λ can be written as a weighted average of the prior mean and the maximum likelihood estimator?

As far I get as the following:-

$E[\lambda | Y] = \frac{\alpha}{\beta} = \frac{n+a}{\sum_{i=1}^{n} Y_i + b}$

Can anyone help me to show how the posterior mean above can be written as a weighted average of the prior mean and the maximum likelihood estimator, $\hat{\lambda}=\bar{y}^{-1}$?

$\endgroup$
2
  • $\begingroup$ Your $P(\lambda|Y_i)$ corresponds to a gamma distribution with $\alpha=n+a$, not $n+a+1$ $\endgroup$ Commented Nov 14, 2018 at 16:23
  • $\begingroup$ @MossMurderer I've changed it. Thank you. I thought it also combination with the likelihood which give Gamma distribution for $\alpha = n+1$. $\endgroup$
    – fhaney
    Commented Nov 14, 2018 at 16:52

1 Answer 1

4
$\begingroup$

Continuing from where you left, $$\begin{align} \bar\lambda_{post}^{-1} &= \frac{\sum_{i}y_i+b}{n+a} \\ &= \frac{n}{n+a}\frac{\sum_{i}y_i}{n} + \frac{a}{n+a}\frac{b}{a} \\ &= \rho\lambda_{mle}^{-1} + (1-\rho)\bar\lambda_{prior}^{-1} \\ \end{align}$$

where $\rho = \frac{n}{n+a}$ and $\lambda_{mle}$, $\bar\lambda_{prior}$ and $\bar\lambda_{post}$ denote the maximum likelihood estimator, prior mean and posterior mean respectively.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.