Durbin-Watson: Autocorrelation Test

Autocorrelation means residuals follow a pattern in sequence — each error influences the next. This is common in time-series or ordered production data.

DW statistic:

• DW ≈ 2.0: ideal — no autocorrelation
• DW < 1.5: positive autocorrelation (errors drift in same direction)
• DW > 2.5: negative autocorrelation (errors alternate direction)

Impact: Autocorrelation inflates R² and makes standard errors unreliable. Consider adding time-lag variables or ordering the data differently.