Mean of squared sum of normal random variables

62 Views Asked by At

Let $X_i \sim N(0, \sigma^2)$ and independent. I would like to show that $\frac{1}{n}\sum\limits_{i=1}^n(\sum\limits_{j=1}^i X_j)^2 \rightarrow \infty$ for $n \rightarrow \infty$ in probability. My problem is that the squared sums aren't independent. Does anyone have an idea which theorem or trick could be useful to solve this.

Many thanks in advance!

1

There are 1 best solutions below

2
On BEST ANSWER

It is enough to show that $(\sum\limits_{i=1}^n X_i) ^{2} \to \infty$ in probability because $Y_n \to \infty$ in probability implies $\frac 1n \sum\limits_{i=1}^{n}Y_i \to \infty$ in probability.

Now let $Z=\frac 1 {\sqrt {n}} {(\sum\limits_{i=1}^n X_i)}$. Then $Z \sim N(0,\sigma^{2})$. Hence $P((\sum\limits_{i=1}^n X_i) ^{2} \leq M)=P(Z^{2} \leq \frac M n) \to 0$ as $ n \to \infty$ for each $M \in (0,\infty)$. This finishes the proof.

Note: $Z_n \to \infty$ in probability iff $Ee^{-Z_n} \to 0$ in probability. Hence $Ee^{-Y_n} \to 0$ in probability Since $e^{-x}$ is a convex function it follows that $Ee^{-\frac 1 n \sum\limits_{i=1}^{n}Y_i} \leq \frac 1 n \sum\limits_{i=1}^{n}Ee^{-Y_i} \to 0$. Hence $\frac 1n \sum\limits_{i=1}^{n}Y_i \to \infty$ in probability.