0
$\begingroup$

I am getting stuck with the following practical statistics problem:

The lifetime duration of one type of lightbulb has exponential distribution. A factory guarantees that the expected lifetime of the lightbulbs they produce is greater than 50 days. In order to achieve high quality production, a worker picks randomly a sample of 40 lightbulbs and notes than in average, they have a lifetime of 53 days.

The factory want to have a 95 percent of probability of not selling if the requirements are not fulfilled

Propose an exact test statistic with simple null hypothesis, what decision should be taken ?

Here is where I got so far:

We have X ~ $\mathcal{E}(\lambda)$ and our null hypothesis is that $H_0$ the null hypothesis is : $$\frac{1}{\lambda} = 50 \space \text{vs} \space \frac{1}{\lambda} > 50$$

Now, I am having some difficulty constructing the exact test, if it was approximate, what I would propose is:

$$T = \dfrac{\sqrt(n)(X_n - E[X])}{\sigma}$$

and since T converges in distribution to $\mathcal{N}(0,1)$, I think that the probability they are talking about is the same as calculating

$P(T > z_{\alpha}) = \alpha$ with $\alpha = 0.5$, so to know what decision should be taken, if $T > z_{\alpha}$ then we reject the null hypothesis and we sell.

But since they are asking for an exact test, I don't know what to do, I think it can be useful the fact that $2\lambda\sum_{i=1}^n X_i$ has distribution $\mathcal{X_{2n}}^2$ but I have no idea how to construct $T$ providing that information. Any help would be greatly appreciated.

$\endgroup$
1
$\begingroup$

Let $n=40$. Under the null $X_1,\dotsc, X_n\stackrel{\text{i.i,d}}{\sim} \text{Exp}(\lambda)$. Then note that $$ \bar{X}=n^{-1}\sum_{i=1}^nX_i\sim\text{Gamma}(n, \lambda n)\tag{0} $$ where we say that $Y\sim \text{Gamma}(p,\lambda)$ (where $p,\lambda>0)$ if $Y$ has the density $$ f_{Y}(y)=\frac{\lambda^{p}}{\Gamma(p)} y^{p-1 }e^{-\lambda y};\quad (y>0). $$ Hence $\bar{X}$ is a test-statistic and we can compute the p-value $p=P(\bar{X}>53)$.

$\endgroup$

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.