Germán Rodríguez
Generalized Linear Models Princeton University

Course Description


This course deals with statistical models for the analysis of quantitative and qualitative data, of the types usually encountered in social science research. The statistical methods studied are the general linear model for quantitative responses (including multiple regression, analysis of variance and analysis of covariance), binomial regression models for binary data (including logistic regression and probit models), models for count data (including Poisson regression and negative binomial models) and models for survival data (focusing on piecewise exponential models fitted via Poisson regression). All of these techniques are covered as special cases of the Generalized Linear Statistical Model, which provides a central unifying statistical framework for the entire course.


The course is taught at an intermediate statistical level. The emphasis is on understanding and applying statistical concepts and techniques, rather than proving theorems. However, the course assumes familiarity with basic concepts in probability theory, statistical estimation and testing theory, and statistical methodology up to multiple regression analysis, at least at the level of a serious introductory course such as WWS507c. Some familiarity with matrix algebra and calculus is necessary. Computer literacy is essential, as we make extensive use of the computer. We recommend using Stata, a general-purpose statistical package available on PCs, Macs and Unix workstations, but students are free to use other software packages such as R or SAS.


Course requirements consist of required readings, six problem sets, and two partial exams, one near the middle and another at the end of the term. Most of the material of the course is covered in formal lectures. A set of lecture notes is distributed, and these can be supplemented with optional readings. The problem sets deal mostly with analysis of small datasets using Stata. The two partial exams emphasize the application of techniques and the interpretation of results. Final grades are calculated as a weighted average of the grades received during the term, using weights of 40% for the problem sets and 30% for each of the two partial exams.

Continue with the List of Lectures