Auto Correlation and Stationarity

Overview

Auto Correlation is a standard tool for detecting whether a time series is stationary or not.

Auto Correlation

The following graph shows a time series that is not stationary.





The resulting auto-correlation shows strong correlations between lagged values and the current value.

Differencing

After differencing the time series above, a time series with no noticeable trends results.

The auto-correlation graph also shows no strong correlations, and no distinct patterns. A stationary time series will show that the auto correlations, if they exist, decay fairly rapidly.

Tools

Auto Correlation App - provides a quick and easy way to chart the auto-correlations of a series.