Why is the law of large numbers defined as the probability of a limit rather than the limit of probability?

235 Views Asked by At

The law of large numbers states for $X_1,...,X_n$ random variables with finite expectation, \begin{equation} \mathbb{P}\bigg(\lim_{n\to\infty}\frac{S_n}{n}=\mu\bigg) = 1 \end{equation} where $S_n = X_1+...+X_n$.

I'm struggling to understand the difference between this definition and

\begin{equation} \lim_{n\to\infty} \mathbb{P}\bigg(\frac{S_n}{n}=\mu\bigg) = 1. \end{equation}

Are they equivilant? Does one imply the other? I know in general that the probability of a limit and the limit of the probability are not equivilant but struggling to understand the reason in this case.

1

There are 1 best solutions below

4
On BEST ANSWER

Essentially you are asking why the probability measure is not continuous. Note that given a sequence of events $A_n$,

$$\lim_{n \to \infty}\mathbb{P}(A_n)=\mathbb{P}\left(\lim_{n \to \infty }A_n\right)$$

holds if $A_n$ is monotonic, i.e. either $A_1\subseteq A_2\subseteq A_3...$ or $A_1\supseteq A_2\supseteq A_3...$

Without monotonicity, continuity need not hold.

Consider $X_1,X_2,X_3,...$ iid $Bern(1/\pi)$. SLLN tells us

$$P\left(\frac{1}{n}\sum_{i=1}^n X_i \overset{n\to \infty}{\to}1/\pi\right)=1.$$

However, since $\sum_{i=1}^n X_i\sim Bin(n,1/\pi)$ is integer-valued,

$$P\left(\frac{1}{n}\sum_{i=1}^n X_i =1/\pi\right)=P\left(\sum_{i=1}^n X_i =n/\pi\right)=0\overset{n\to \infty}{\not\to}1.$$