Limit of integral/expectation

136 Views Asked by At

Let $\xi_1,\xi_2,\ldots,\xi_n$ be an independent sample of Exp($1$) random variables with joint density, $f(x_1,\ldots,x_n) = e^{-(x_1+\cdots+x_n)}$. If $g$ is a bounded continuous function, compute the following limit,

$$ \lim\limits_{n\rightarrow\infty} \int_{\mathbb{R}^n_+}g \left( \frac{x_1+ \cdots +x_n}{n} \right)f(x_1,\ldots,x_n) \, dx_1\cdots dx_n $$

where $\mathbb{R}_+^n = (0,\infty)^n$. My guess is that it will converge to $g(1)$. My reasoning is that we can view the integral as the expectation, $\operatorname{E}[g(\bar{X}_n)]$. By the strong law of large numbers and the coninuous mapping theorem, $\bar{X}_n \xrightarrow{\text{a.s.}} 1$ and $g(\bar{X}_n) \xrightarrow{\text{a.s.}} g(1)$. Since $g$ is bounded, we are permitted to bring the limit inside the integral,

$$ \lim_{n\rightarrow\infty} \operatorname{E}[g(\bar{X}_n)] = \operatorname{E} \left[ \lim_{n\rightarrow\infty} g(\bar{X}_n) \right] = g(1) $$

This is very handwavy, and probably not true though. Can someone lead me in the right direction?

1

There are 1 best solutions below

2
On BEST ANSWER

As Shalop mentioned, this is not hand wavy since you checked that you can use the dominated convergence theorem in the probabilistic setting. The sequence $\left(g\left(\overline{X_n}\right)\right)_{n\geqslant 1}$ is bounded by $\sup_{x\in\mathbb R}\left\lvert g(x)\right\rvert$ and converges almost surely to $g(1)$.