Pros and cons

• Diagnostic methods based on residuals are a very useful to identify:

  • Problems with distributional assumptions.

  • Problems with the assumed structure of the model (in terms of the selection of the explanatory variables and their form).

  • Groups of observations for which a model’s predictions are biased.

• It presents the following limitations:

  • Interpretation of the patterns seen in graphs may not be straightforward.

  • It’s not be immediately obvious which element of the model may have to be changed.