2005 Fall STA 216-01
Bulletin Course Description Likelihood-based and Bayesian inference of binomial, ordinal, and Poisson regression models, and the relation of these models to item response theory and other psychometric models. Focus on latent variable interpretations of categorical variables, computational techniques of estimating posterior distributions on model parameters, and Bayesian and likelihood approaches to case analyses and goodness-of-fit criterion. Theory and practice of modern regression modeling within the unifying context of generalized linear models. A brief review of hierarchical linear models. Students expected to use several software packages and to customize functions in these packages to perform applied analyses. Instructor: Dunson
(Instructor named in bulletin description above may not be current. For current instructor, see listing below.)
Title GENERALIZED LIN MODELS Department STA Course Number 2005 Fall 216 Section Number 01 Primary Instructor Dunson,David Prerequisites Prerequisite: Statistics 213 and 244 or consent of instructor.
Prerequisites
STA 213, STA 215, MTH 136, and STA 244 or permission of the instructor.
Synopsis of course content
Some review of multiple linear regression, theory and practice. Theory of likelihood-based inference for generalized linear models (GLMs). Emphasis on binary and ordinal data models, with examples in item response theory. Some coverage of log-linear models. Data analysis: model fitting, model choice, and residuals-based diagnostics. Bayesian theories of inference in GLMs.
Additional Information
This course is meant for graduate students or advanced undergraduate students.
