0
$\begingroup$

Let $(X_n)_n$ be independent and identically distributed random variables with expected value $\mu$, variance $\sigma^2$ and $E(X_n^4)<\infty$. How can I show that

$$\lim_{n \to \infty}P\left(\left|{1 \over n}\sum_{i=1}^n\left(X_i-{{X_1+X_2+\cdot \cdot\cdot + X_n} \over n }\right)^2 - \sigma^2 \right| >\epsilon\right)=0$$ for all $\epsilon > 0$.

Some help is much appreciated.

$\endgroup$
1
$\begingroup$

Using the Strong Law of Large Numbers, we can actually say that $$ \frac1n \sum_{i = 1}^n \bigg( X_i - \frac{X_1 + \cdots + X_n}{n} \bigg)^2 $$ converges almost surely to $\sigma^2$. This is stronger than convergence in probability. To see this, for the $i$-th summand we can expand to get $$ X_i^2 - 2 \bar S_n X_i + \bar S_n^2 $$ Here $\bar S_n = \frac1n (X_1 + \cdots + X_n)$. For the first term, $$ \frac1n \sum_{i = 1}^n X_i^2 $$ converges almost surely to $E(X_1^2)$ by the Strong law. For the second term, $$ - 2 \frac1n \sum_{i =1}^n X_i \bar S_n = - 2 (\bar S_n)^2 $$ converges to $- 2 \mu^2$ as $n \to \infty$, again by the strong law. Similarly, the last term $\frac1n \cdot n \bar S_n^2 = \bar S_n^2$ converges to $\mu^2$. Putting it all together, the almost-sure limit is $$ E(X_1^2) - \mu^2 = \sigma^2 \, . $$

I would point out that all we used was the fact that $E(X_1^2) < \infty$ to invoke the Strong Law for $\frac1n \sum_{i = 1}^n X_i^2$.

$\endgroup$

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.