Chapter 19 Residual-Diagnostics Plots

Learning Objectives:

This section presents graphical methods for a detailed examination of model performance at both overall and instance-specific levels.

• Residuals can be utilized to:
  • Identify potentially problematic instances. This can help define which factors contribute most significantly to prediction errors.
  • Detect any systematic deviations from the expected behavior that could be due to:
    • The omission of explanatory variables
    • The inclusion of a variable in an incorrect functional form.
  • Identify the largest prediction errors, irrespective of the overall performance of a predictive model.