A stationary time series is a time series whose probabilistic behavior does not change with time. A stochastic
process
has strict stationarity if the joint
distribution of
|
(1)
|
is the same as that of for every
, every choice of times, and every shift
.
A process with finite second moments has weak stationarity, also called covariance stationarity, if
|
(2)
| |||
|
(3)
|
where
denotes the expectation value of
. Thus, its mean is constant
and its covariance depends only on the lag
. Strict stationarity
does not in general imply weak stationarity
unless the necessary moments exist, and weak
stationarity does not in general imply strict
stationarity.