Session 3: Modelling Binary and Categorical Outcomes
Note
Date: Thursday, 30 April 2026, 16:00–19:00
Business question
Can we predict which firms are likely to exit the market in the next year?
Learning goals
- Understand when and why OLS breaks down for binary outcomes
- Run and interpret logistic regression with
glm() - Translate log-odds into meaningful business insights (predicted probabilities, marginal effects)
Dataset
bisnode-firms — ~17,500 Hungarian firms in manufacturing and services (Békés & Kézdi 2021). Key variables: firm exit (binary), log sales, log employees, profit margin, firm age, liquidity, ownership structure.
Session outline
| Time | Activity |
|---|---|
| 16:00–16:30 | Debrief: Take-home Task 2 |
| 16:30–17:10 | Input: why OLS fails for binary outcomes; logistic regression intuition |
| 17:10–17:20 | Break |
| 17:20–18:05 | Live demo |
| 18:05–18:45 | In-session exercise |
| 18:45–19:00 | Debrief + Quarto skill: inline R code |
Materials
| File | Description |
|---|---|
| Slides | Lecture slides — open in browser, press F for fullscreen |
| Lecture notes & demo | Live coding document built during the session |
| Exercise sheet | In-session exercise — predicting firm exit with logistic regression |
| Session outline | Detailed session plan with timing and teaching notes |