consider a sequence $\{X_j\}$ of iid random variables where $X_j$ is in $L_1$ for each j

527 Views Asked by At

Consider a sequence $\{X_j\}_{j\geq1}$ of iid random variables where $X_j$ is in $L_1$ for each $j$. Suppose that the random variable $\frac{1}{\sqrt{n}} \sum_{j=1}^{n}(X_j-a)$ converges in law to the random variable $W$.

Show that $\lim_{n\rightarrow \infty}\frac{1}{n} \sum_{j=1}^{n}X_j=a$ almost surely.

I know that, by the law of large number, $\frac{1}{n} \sum_{j=1}^{n}X_j$ converges almost surely to the expectation of $X_1$. But I don't know how to prove that $a$ is the expectation of $X_1$.

How can I solve it? Thank you for your help!

1

There are 1 best solutions below

0
On BEST ANSWER

Set

$$S_n := \sum_{j=1}^n (X_j-a).$$

By assumption, $S_n/\sqrt{n} \to W$ in distribution for some real-variabled random variable $W$. Fix $\delta, \epsilon>0$, and choose $R>0$ such that $\mathbb{P}(|W| \geq R) \leq \epsilon$. Since

$$\mathbb{P} \left( \left| \frac{S_n}{n} \right| > \delta \right) \leq \mathbb{P} \left( \left| \frac{S_n}{\sqrt{n}} \right| \geq R \right)$$

for $n \geq N=N(R)$ sufficently large, we find by the Portmanteau theorem

$$\limsup_{n \to \infty} \mathbb{P} \left( \left| \frac{S_n}{n} \right| > \delta \right) \leq \mathbb{P}(|W| \geq R) \leq \epsilon.$$

This shows that $S_n/n \to 0$ in probability. As

$$\mathbb{P} \left( \left| \frac{1}{n} \left( \sum_{j=1}^n X_j \right) - a \right|>\delta \right) = \mathbb{P} \left( \left| \frac{S_n}{n} \right|>\delta \right)$$

this means that $\frac{1}{n} \sum_{j=1}^n X_j \to a$ in probability. On the other hand, we have by the central limit theorem

$$\frac{1}{n} \sum_{j=1}^n X_j \to \mathbb{E}(X_1)$$

almost surely and hence in probability. Since limits in probability are unique (up to a null set), we get

$$a=\mathbb{E}(X_1).$$

Remark: The first part of this proof actually shows the following statement:

If $(Y_n)_{n \in \mathbb{N}}$ is a sequence which converges in distribution, then $\frac{Y_n}{\sqrt{n}} \to 0$ in probability.