STATS 4C03/6C03, Fall 2018
GENERALIZED LINEAR MODELS
While the normal linear regression model has been studied in many courses (most prominently in STATS 3A03), there are many statistical datasets where the distribution of the response variable is non-normal. For instance, the distribution of counts is more likely to be binomial or Poisson, while the distribution of failure times for products or survival times of patients with a certain condition is more likely to be gamma or lognormal. The regression models appropriate to these situations are called generalized linear models. The model structures, including, will be presented along with the properties and estimation methods. Extensive use of R will be made to fit the models to a variety of datasets. The course prerequisites are STATS 3A03 and STATS 3D03.
INSTRUCTOR: B. Bolker
Normal linear model, exponential family, iteratively-reweighted least squares, logistic regression, Poisson regression and log-linear models, other families of GLM’s, analysis of deviance and model checking, residual analysis.
Three lectures; one term
Prerequisite(s): STATS 3A03 and STATS 3D03
PLEASE REFER TO MOSAIC FOR THE MOST UP-TO-DATE INFORMATION ON TIMES AND ROOMS