9 B2. Limited Dependent Variables — Overview
9.1 About
This section covers Limited Dependent Variable (LDV) models — regression models where the outcome \(y\) is discrete, censored, or bounded. When \(y\) is binary or count-valued, OLS predictions fall outside the feasible range and errors are heteroskedastic by construction. MLE-based alternatives are required.
Topics covered in this block: binary choice (Logit, Probit), multinomial choice, count data (Poisson), censoring and sample selection (Tobit, Heckman).
9.2 Lecture Notes
9.3 Slides
Click on the slide and use the keyboard arrows to navigate.
Note: Slides and notes are pending translation to English.
9.4 Overview
This block covers limited dependent variable (LDV) models — regression models where the outcome variable \(y\) is discrete, censored, or otherwise restricted in its range. These models arise frequently in microeconometrics and applied economics.
Why standard OLS fails
When \(y\) is binary (0/1) or non-negative integer-valued, OLS predictions may fall outside the feasible range, error terms are heteroskedastic by construction, and predicted probabilities can be negative or exceed 1. MLE-based alternatives are required.
Sections
- B3. Binary Choice Models — LPM, Logit, Probit, marginal effects, random utility.
- B4. Multinomial Choice Models — Multinomial Logit, IIA, Conditional Logit, Ordered models.
- B5. Count Data Models — Poisson regression, overdispersion, Negative Binomial.
- B6. Censoring, Truncation & Sample Selection — Tobit, Heckman selection model.