Auto-correlation in the residuals can be masked/suppressed by Pulses/Seasonal Pulses or the resultant of omitted Level/Step shifts or Local Time Trends. Auto-correlation in the residuals may be the result of parameters changing over time. Auto-correlation in the residuals may be the result of omitted (important) lag structures in the user specified cause variables. Auto-correlation in the residuals may be the result of an omitted (important) cause variable either stochastic or deterministic. Correct strategies to deal with auto-correlation depend on finding/detecting which of the above is relevant. Simple approaches assuming the nature of the problem/opportunity often miss the boat.

Auto-correlation in the residuals can be masked/suppressed by Pulses/Seasonal Pulses or the resultant of omitted Level/Step shifts or Local Time Trends. Auto-correlation in the residuals may be the result of parameters changing over time. Auto-correlation in the residuals may be the result of omitted (important) lag structures in the user specified cause variables. Auto-correlation in the residuals may be the result of an omitted (lurking) cause variable either stochastic or deterministic. Correct strategies to deal with auto-correlation depend on finding/detecting which of the above is relevant. Simple approaches assuming the nature of the problem/opportunity often miss the boat.