Autocorrelation
The correlation of a variable with itself across different points in time or space.
What is Autocorrelation?
Also known as serial correlation, autocorrelation occurs when the value of a variable at one time point is highly dependent on its value at previous time points. In the context of regression, it often refers to the residuals being correlated with each other, which violates the assumption of independent errors.
Why Autocorrelation Matters
In time series data, unaddressed autocorrelation makes standard errors too small and inflates t-statistics, leading researchers to conclude a relationship is statistically significant when it is not.
Example
A meteorologist studying daily temperatures finds that today's temperature is highly correlated with yesterday's temperature, requiring time-series specific modeling to handle the autocorrelation.
Common Mistakes
- Using standard Ordinary Least Squares (OLS) regression on time series data without testing for autocorrelated residuals (e.g., using the Durbin-Watson test).
- Assuming autocorrelation only happens in time series, ignoring spatial autocorrelation in geographic data.