For a random variable x,
Statistical convergence can be defined such as
for any ε > 0.
For normal data
|
: Student-t Statistic |
For non-normal data with unknown distribution
|
: Tchebycheff inequality |
For a stationary random variable x,
where, d is the confidence interval size, x is the sample mean, sx is the standard deviation, and N is the number of data.
For a 95% confidence level:
If x is normal, c = 2.0 by using the Student-t statistics,
If x is not normal with unknown distribution, c = 4.5 by using the Tchebycheff inequality.
For a random process {x(t)}*, mean value μx(t) and autocorrelation function Rx(t,t+τ) are given by
* the symbol { } is used to denote an ensemble of sample function.
Stationary: μx(t) and Rx(t,t+τ) do not vary as time t varies, i.e., μx(t) = μx and Rx(t,t+τ) = Rx(τ).
Weakly stationary
Strongly stationary
Self-stationary: for the kth same function, the sample properties μx(t,k) and Rx(t,t+τ,k) do not vary.