Estimating a highest probability density (HPD) interval

93 Views Asked by At

QUESTION: (I translated it into English so you might find somewhat awkward..)

A company wants to get information about the lifespan of certain machines. Let's say that the lifespan of the machines follows an exponential distribution which is $p(y|\theta)=\theta e^{-\theta y}(\,y>0)$

After examining 50 machines, the average lifespan of the machines was 756 hours. Find a 95% HPD credible interval for $\theta$.(Use a non informative prior $p(\theta)=\theta^{-1}$


Hello. I'm studying Bayesian statistics with some textbook and I came across this question. I guess I'm having trouble understanding this part => 'After examining 50 machines, the average lifespan of the machines was 756 hours'. Does this mean that there are 50 samples every which follows an exponential distribution..?

If I think of it like that, the likelihood would be $p(y_{1},..,y_{50}|\theta) = \theta^{50}e^{-\theta\sum_{i=1}^{50}\large y_{i}}$ and by using a given prior we can get the posterior distribution $p(\theta|y_{1},..,y_{50}) \propto \theta^{49}e^{-\theta\sum_{i=1}^{50}\large y_{i}}$ which is Gamma(50, $\sum_{i=1}^{50}\large y_{i}$)

Now I have everything except $\sum_{i=1}^{50}\large y_{i}$. I guess I can get something from 'average lifespan was 756 hours' but isn't this average lifespan is about parameter $\theta$? How can I derive something that has to do with $\sum_{i=1}^{50}\large y_{i}$

Any help is appreciated.

1

There are 1 best solutions below

0
On BEST ANSWER

Does this mean that there are 50 samples every which follows an exponential distribution..?

Yes. All samples taken are assumed to come from an exponential distribution.

You have the sample mean of the $y_i$: $1/50 \sum_i^{50}y_i = 756$ so now you can find the summation you are looking for to compute your posterior.