Part 5 | Multivariate GLM
> models with multiple predictorsIn Part 4 we developed the basics of linear models, extending the hypothesis testing framework to slopes. In Part 5 we extend this same framework one step further. We'll develop models of relationships that vary across groups, effects that depend on context, and unobserved factors that confound your analysis. You'll learn advanced techniques like numerical controls, categorical controls, interactions, and model selection that let you handle these complexities systematically.
Part 5.1 | Categorical/Numerical
Use fixed effects to control for unobserved group differences. Interpret coefficients in models with multiple categorical predictors.
Livestream 5.1
Class recording from Fall 2025
Exercise 5.1 // Gender Wage Gap
Testing whether wages are different by gender
Homework 5.1 // Categorical Controls
Due on Sunday April 12 at 11:59PM on Gradescope.
Part 5.2 | Interactions
Test whether the relationship between variables varies by group.
Livestream 5.2
Class recording
Exercise 5.2 // Gender Wage Gap
Is the return to education different by gender?
No notebook
Homework 5.2 // Interactions
Due on Sunday April 12 at 11:59PM on Gradescope.
Part 5.3 | Numerical Controls
Control for related variables.
Livestream 5.3
Class recording from Fall 2025
Homework 5.3 // Numerical Controls
Due on Sunday April 19 at 11:59PM on Gradescope.
Part 5.4 | Causation, Controls, and Model Selection
Why controls matter, how to compare models, and how to choose the right one.
Livestream 5.4
Class recording
Concept 5.4 // Causation, Controls, and Model Selection
Causation vs. correlation, R² and F-tests, choosing the right model
Exercise 5.4 // Model Comparison with the F-Test
Testing whether controls improve the model.
Homework 5.4 // Causation and Model Selection
Due on Sunday April 19 at 11:59PM on Gradescope.
MiniExam
MiniExam 5 will be held on Thursday April 16 and will cover everything in Part 5: numerical controls, categorical controls, interactions.