1. Forecasting models based on decomposition methods
2. Adding ETS and ARIMA filters
- 2.1 Exponential Smoothing
- 2.2 Identifying Stationarity
- 2.3 ARIMA models
- 2.4 Decomposition & ETS/ARIMA
3. Adding exogenous variables to the model
- 3.1 Linear Regression Models
- 3.2 Build dynamic regressions
- 3.3 Analyze the model's performance
- 3.4 Choose the best variables for the model
4. Forecasting at scale
- 4.1 Understanding the challenges and approaches when dealing with multiple time series
- 4.2 Applying hierarchical and reconciliation models
- 4.3 Efficient forecasting workflow
- 4.4 Evaluation of the global performance

Identifying Stationarity

Modified

February 13, 2026

A time series is said to be stationary if its statistical properties (mean and variance) are constant over time.

Stationarity is important because many time series forecasting methods assume that the underlying data is stationary.
Non-stationary data can lead to inaccurate forecasts and misleading interpretations.

There are two main types of stationarity:
- Strict Stationarity: The joint distribution of any set of observations is the same regardless of the time at which they are observed.
- Weak Stationarity: The mean, variance, and autocovariance of the series are constant over time.
In practice, we often focus on weak stationarity because it is easier to test and work with.