Skip to content
ML Academy

Introduced on day 21

Stationarity

A series is (weakly) stationary when its mean, variance and autocorrelation do not depend on time; trend, level shifts and a fixed seasonal pattern all violate it.

In detail

Stationarity is a property of the process, so it is judged on the series after the transformations you plan to apply (log, differencing, seasonal adjustment). ARIMA-type models assume it; so does any method that treats the past as informative about the future. A rolling mean and variance give a quick visual check; KPSS (null: stationary) and ADF (null: unit root) are formal tests that ask opposite questions and can disagree. A stationary series is not necessarily predictable: stable moments say nothing about forecast skill.