5 · Multiple regression

Several predictors, and everything that can go wrong

The core of the course. Six lecture decks and ten exercises covering multiple regression itself and the four assumptions that most often fail in applied work: linearity, no omitted variables, no collinearity, and constant error variance.

Slides

Deck	Covers
🖥 Multiple linear regression	Several predictors at once; holding other things equal in practice
🖥 Linearity	Diagnosing non-linearity; fixing it with transformations
🖥 Omitted variable bias	What leaving a variable out does to the estimates you keep
🖥 Collinearity	When predictors carry the same information; VIF and diagnostics
🖥 Heteroscedasticity	Non-constant error variance and robust standard errors
🖥 Hedonic prices model	A full applied example decomposing prices into characteristics

Exercises

Multiple regression and linearity

Exercise	What you practise
📋 AE04-01 Multiple linear regression	Fitting and reading a model with several predictors
📋 AE04-02 Linearity	Residual plots, log transformations, interaction terms, marginal effects
📋 AE04-03 Hedonic land prices	Homework: a hedonic land price model end to end
📋 AE04-04 OVB from the slides	Homework: reproduce the omitted-variable-bias examples

📦 Bundle: ae04-multiple-regression-part-1

Omitted variable bias and heteroscedasticity

Exercise	What you practise
📋 AE05-01 OVB and the wage equation	Omitted variable bias in a classic wage regression, plus the homework section
📋 AE05-02 OVB from the slides	Homework: further OVB practice
📋 AE05-03 Hedonic prices & heteroscedasticity	A complete analysis with heteroscedasticity tests and robust standard errors

📦 Bundle: ae05-multiple-regression-part-2

Transformations, interactions, heterogeneity

Exercise	What you practise
📋 AE06-01 Transformations & interactions toolbox	The full toolbox for transformed and interacted regressors
📋 AE06-02 Interactions with continuous variables	Interpreting continuous-by-continuous interactions
📋 AE06-03 Simpson’s paradox	Unobserved heterogeneity reversing a relationship

📦 Bundle: ae06-multiple-regression-part-3

Going deeper

Read

• 📖 Chapter 2, Regression, in (Angrist and Pischke 2014) — find it in the JLU library
• 📖 IMS Ch. 8 — Linear regression with multiple predictors

Watch

• 📺 Regression Part I: Call in the CIA
• 📺 Regression Part II: Theory

Practise

• Open Intro lab: Multiple regression

Where the course goes next

← 4 · Simple regression · → 6 · Panel regression

References

Angrist, Joshua D., and Jörn-Steffen Pischke. 2014. Mastering’metrics: The Path from Cause to Effect. Princeton University Press.