Strict Stationarity

Definition

$\{X_t, t \in T\}$ is strictly stationary if for any $n \geq 1$, times $t_1, \ldots, t_n$, and lag $h$:

$$\Pr(X_{t_1} \leq x_1, \ldots, X_{t_n} \leq x_n) = \Pr(X_{t_1+h} \leq x_1, \ldots, X_{t_n+h} \leq x_n)$$

The entire joint distribution is invariant under time shifts.

Consequences (if moments exist)

  • Constant mean: $E(X_t) = \mu$
  • Constant variance: $\text{Var}(X_t) = \sigma^2$
  • ACVF depends only on lag

Example

i.i.d. random variables are always strictly stationary.