DataTerminology.comNEW
A time series property where mean, variance, and autocovariance remain constant over time.
DESCRIPTION
Stationarity underpins many time series methods by assuming stable statistical properties. In weak (covariance) stationarity, mean and variance are constant and autocovariance depends only on lag. Non-stationary series often require differencing, detrending, or transformations. Tests like ADF and KPSS assess stationarity from complementary perspectives. Achieving stationarity enables reliable inference, forecasting, and model diagnostics. Seasonal stationarity may require seasonal differencing.
ALTERNATIVE DEFINITIONS
Penn State Eberly College of Science
A time series is said to be stationary if its statistical properties do not change over time.
Concise emphasis on invariance of statistical characteristics.
NIST Engineering Statistics Handbook
A stationary process has a constant mean, constant variance, and an autocovariance that does not depend on time.
Specifies all three core conditions for weak stationarity.
RELATED TERMS
Unit Root Differencing ARIMA Random Walk Augmented Dickey–Fuller Test Autocorrelation White Noise Covariance Stationarity Ergodicity Variance Stabilization
COMPETING TERMS
Non-stationarity
KNOWLEDGE AREAS
TARGET AUDIENCE
Data Scientist Data Analyst AI Engineer
EXTERNAL SOURCES
- Wikipedia - Stationary process
- FPP3 (Hyndman & Athanasopoulos) - Stationarity primer and examples
- statsmodels - ADF test documentation
TOP QUESTIONS
Why does Stationarity matter in time series modeling?
Many models assume stationarity to ensure stable parameter estimates and valid inference. Non-stationary data can cause spurious relationships and unreliable forecasts.
Category: Relevancy
How can I test Stationarity?
Use ADF, KPSS, and PP tests for complementary evidence. Visual inspection of trend, variance stability, and autocorrelation patterns also guides assessment.
Category: How-to
What transforms help achieve Stationarity?
Differencing (regular/seasonal), detrending, and variance-stabilizing transforms like log or Box-Cox are common. Choose minimally invasive transformations that stabilize mean and variance.
Category: How-to
What is the difference between strict and weak Stationarity?
Strict stationarity requires all joint distributions be time-invariant. Weak (covariance) stationarity requires only constant mean, variance, and lag-dependent autocovariance.
Category: Concept
How does Seasonality affect Stationarity?
Seasonal patterns violate stationarity; apply seasonal differencing or model seasonal components to achieve stationary residuals.
Category: Concept
What is a common mistake regarding Stationarity?
Relying on a single test like ADF without residual diagnostics. Combine tests with plots and domain context to avoid misclassification.
Category: Mistakes
Is differencing always necessary for Stationarity?
No. Some series become stationary with detrending or transformation alone. Over-differencing can introduce unnecessary noise and invertibility issues.
Category: Opinion
How does Stationarity relate to ARIMA?
The “I” in ARIMA stands for differencing to achieve stationarity. AR and MA terms then model the stationary structure.
Category: Relation
Does Stationarity guarantee good forecasts?
No. It’s necessary for many models but not sufficient. Proper model specification, diagnostics, and exogenous factors still determine forecast quality.
Category: Misunderstanding