Teaching
Advanced Social Statistics
Sociology 504 is the second course in the graduate social statistics sequence. It builds on Soc500 and the Princeton Sociology Summer Methods Camp, with more intensive treatment of maximum likelihood estimation, generalized linear models, and advanced tools for empirical social science.
The 2017 version was taught as SOC 401/504 and used a two-part structure. The first half focused on core GLM and likelihood material; the second half moved through modules on latent variables and missing data, causal inference, regularization, and hierarchical models.
Course Notes
The course design draws heavily on shared teaching materials from Gary King's Gov2001 and from Matt Blackwell, Justin Grimmer, Erin Hartman, Teppei Yamamoto, and others. Rebecca Johnson and Ian Lundberg served as teaching assistants for the 2017 course, and their precepts are included below.
Soc504 differs from Soc500 in pacing and format. Lectures in the first half of the semester move fluidly across class meetings, while the later modules are more sharply divided into multi-week topic blocks. The precepts and problem sets are central to making the material concrete.
Replication Project Materials
The course centers on a replication and extension project in which student pairs select a recent paper, reproduce its main results, and build an extension into a final paper.
2017 Lectures and Precepts
February 6
Lecture 1: Introduction
February 8
Lecture 2: Basics
February 9
Precept 1: Review of Probability, Simulations and Data Manipulation
Rebecca Johnson
February 13-20
Lecture 3: Maximum Likelihood Estimation
February 16
Precept 2: Likelihood Inference
Ian Lundberg
February 23
Precept 3: Numerical Optimization and Simulation
Rebecca Johnson
February 22-March 15
Lecture 4: Generalized Linear Models
March 2
Precept 4: Binary and Lognormal GLMs
Ian Lundberg
March 8
Precept 5: Binary and Ordinal Outcomes
Rebecca Johnson
March 16
Precept 6: Duration and Count Data
Ian Lundberg
March 20-24
Spring Break
March 27-April 5
Lecture 5: Latent Variables, EM and Missing Data
Mixture models, expectation maximization, missing data, and multiple imputation.
March 28
Precept 7: Mixture Models and EM
Rebecca Johnson
April 6
Precept 8: Missing Data
Ian Lundberg
April 10-19
Lecture 6: More Causal Inference
Model dependence, matching, propensity scores, and mediation.
April 12
Precept 9: Model Dependence and Matching
Rebecca Johnson
April 20
Precept 10: Matching, Mediation and Dynamic Treatments
Ian Lundberg
April 24-26
Lecture 7: Regularization
Regularization, the eight schools example, and hierarchical models.
May 1