Please read section 2.2.4 Best Predictors and Best Linear Predictors in Foundations of Agnostic Statistics.
- Pages 75—82
- The section titled Best Predictors and Best Linear Predictors
This is a longer section of reading, so hear is a reading guide:
- Pages 75—76 introduce the concept of the Best Linear Predictor.
- Pages 77—78 uses calculus to derive a formula for the BLP. Because the function is linear in the terms, the derivative of the function is simplifies really nicely.
- Page 79 introduces properties of deviations from the BLP. Take note about what is similar (and what is different) compared to deviations from the CEF. Why does the BLP provide fewer guarantees about deviations from the function than the CEF?
- Pages 79—81 go through the arduous/tedious process of estimating a BLP by hand, given a specific functional form (i.e. formula) for the variables. Follow through this example; we'll talk about it in a lecture momentarily.