Given $\bar{X_n}$ is mean of random sample with size $n$ from Gamma distribution with parameter $\alpha=\mu$ and $\beta=1$. I wanna find the limiting distribution of $\frac{\sqrt{n}\left(\bar{X_n}-\mu\right)}{\sqrt{\bar{X_n}}}$.

Can I use the Central Limit Theorem for doing this? Because it is $\sqrt{\bar{X_n}}$ in the denominator, not $\sqrt{S^2}$.

I am doing this problem with this method :

Consider that the Moment Generating Function (MGF) of $\bar{X_n}$ is $M_\bar{x}\left(t\right)=\left(1-\frac{t}{n}\right)^{-n\mu}$ for $t<n$. So that, the pdf of $\bar{X_n}$ is

$f_n\left(\bar{x}\right)=\frac{1}{\Gamma\left(n\mu\right)\left(\frac{1}{n}\right)^{n\mu}}\bar{x}^{n\mu-1}e^{-n\bar{x}}$ for $0<\bar{x}<\infty$

From the MGF, we know that the mean of $\bar{X_n}$ is $\mu$ and the variance of $\bar{X_n}$ is $\frac{\mu}{n^2}$.

Then, consider the transformation $Y_n=\frac{\sqrt{n}\left(\bar{X_n}-\mu\right)}{\sqrt{\bar{X_n}}}$ so that,


Then I get stuck to define $\bar{x}$ in function of $y$ to find pdf and CDF of $Y$

Any simpler method? Or I do mistake when I define the transformation? Any ideas?


Just write

\begin{align} \frac{\sqrt{n} (\bar{X} - \mu)}{\sqrt{\bar{X}}} &= \frac{\sqrt{n} (\bar{X} - \mu)}{\sqrt{\mu}} \cdot \sqrt{\frac{\mu}{\bar{X}}} . \end{align}

By the central limit theorem the first term has a limiting normal$(0, 1)$ distribution and we also know $\bar{X} \to \mu$ with probability one by the strong law of large numbers, so the second factor converges to one almost surely. Then appealing to Slutsky's theorem the entire thing goes to a normal$(0, 1)$.

  • $\begingroup$ OH I GOT IT. Thank you very much for the concept. ^^ $\endgroup$
    – MusMeong
    Jan 29 '16 at 15:31

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.