Asymptotic distribution of $\left(1-\frac{1}{n}\right)^{n\bar{X}_n}$

83 Views Asked by At

Suppose $X_1,X_2, \cdots$ are i.i.d. observations from a $Poisson(\lambda)$ distribution. Define $\bar{X}_n=\sum_{i=1}^nX_i/n$. What will be the asymptotic distribution of $\left(1-\frac{1}{n}\right)^{n\bar{X}_n}$?

I have solved this problem, however I need to use the facts that $\left(1-\frac{1}{n}\right)^{n\bar{X}_n}$ and $e^{-\bar{X}_n}$ are the UMVUE and MLE for $e^{-\lambda}$ respectively. I don't how to use this fact. Please tell me how to do this problem using this interesting fact.

1

There are 1 best solutions below

0
On BEST ANSWER

No idea why these (not so) "interesting facts" should be used (in this context) since, considering the random variables $Y_n=\left(1-\frac{1}{n}\right)^{n\bar{X}_n}$, the fact that $\mathrm e^{-(n+1)/n^2}\leqslant1-1/n\leqslant\mathrm e^{-1/n}$ for every $n\geqslant2$ yields readily $$ \mathrm e^{-(1+1/n)\bar X_n}\leqslant Y_n\leqslant\mathrm e^{-\bar X_n}. $$ Thus, the almost sure convergence $\bar X_n\to\lambda$ implies that $Y_n\to\mathrm e^{-\lambda}$ almost surely (and in probability and in distribution and in every $L^p$, $0\lt p\lt\infty$).