Model understanding through predictions (Model Evaluation by Prediction Difference)
Understanding a model by looking at its predictions, not just its parameters.
This has a big advantage: every type of predictive model
makes predictions (otherwise what use would it be?) so we can use
the same set of techniques to understand any type of predictive
model.
Model understanding through residuals (Residuals Analysis)
Residuals are powerful because they allow us to use models to remove striking patterns so we can study the subtler trends that remain.