E.1 Law of large numbers
Suppose we observe data D=\{x_1, \ldots, x_n\} with each x_i \sim F .
By the strong law of large numbers the empirical distribution \hat{F}_n based on data D=\{x_1, \ldots, x_n\} converges to the true underlying distribution F as n \rightarrow \infty almost surely:
\hat{F}_n\overset{a. s.}{\to} F
The Glivenko–Cantelli asserts that the convergence is uniform. Since the strong law implies the weak law we also have convergence in probability:
\hat{F}_n\overset{P}{\to} F
Correspondingly, for n \rightarrow \infty the average \text{E}_{\hat{F}_n}(h(x)) = \frac{1}{n} \sum_{i=1}^n h(x_i) converges to the expectation \text{E}_{F}(h(x)) .