3.12 Chapter 3 Major Takeaways

1. All models are wrong, but some are useful and fair for our goals (prediction or description).

2. We want a model with small residuals or a low number of prediction errors.

3. To determine if a model is useful and fair, we study what is left over (the residuals).

4. We use models to describe phenomena by interpreting estimated coefficients. But make sure you are talking about the average or predicted outcome! If you have multiple variables, you are keeping all others fixed (if possible).

5. We also use models to make predictions, but be careful about extrapolating: predicting outside the observed range of our predictor (\(X\)) variables.