Durbin-Watson test
The Durbin-Watson test detects first-order autocorrelation in regression residuals, producing a statistic between 0 and 4. A value near 2 indicates no autocorrelation; near 0 signals positive, near 4 negative serial correlation.
FrameworkClassical linear regression model
See it move
The Durbin-Watson statistic sums the squared differences between consecutive residuals and divides by the total sum of squared residuals, always landing between 0 and 4. A value near 2 means no first-order autocorrelation; values drifting toward 0 signal positive autocorrelation, and values drifting toward 4 signal negative autocorrelation. Values between 1.5 and 2.5 are generally considered acceptable.
The formula
Variables
- OLS residual at observation t
- OLS residual at the preceding observation
- Total number of time-series observations
DW ranges from 0 to 4. DW ≈ 2 indicates no first-order autocorrelation; DW < 2 positive autocorrelation; DW > 2 negative autocorrelation. Values between 1.5 and 2.5 are generally considered acceptable.