STA721 (Formerly STA 244): Linear Models

Course Home | Computing Links | Course Calendar |Sakai| DSS Home | Duke

Meeting Times: Old Chem 101 Tues-Thur 1:15-2:30

Instructor: Merlise Clyde

Teaching Assistant: Nick Jarrett (nicholas.jarrett@duke.edu)

Course Description

This course investigates the essential concepts of linear models from both Bayesian and classical viewpoints, using a coordinate free approach where possible. Topics include: simple linear and multiple regression, parameter estimation and interpretation, distribution theory for ANOVA and testing, variable transformations, prediction, model diagnostics, variable selection, Bayes factors and model selection, and Bayesian model averaging. and Bayesian hierarchical linear models. Selected topics in Markov chain Monte Carlo simulation will be introduced as required. Extensions to multivariate models and nonparametric models if time permits.

Prerequisite: Statistics 213 and Linear Algebra (Math 104), Corequisite: Statistics 290

Grading will be based on Homeworks, Midterms (In-class/Takehome) and Final.

References:

The following books will be useful for the course:

• Plane Answers to Complex Questions: The Theory of Linear Models- Ronald Christensen (2011) 4th Edition Springer-Verlag, NY (required)
• A First Course in Bayesian Statistical Methods Peter Hoff (2009) 1st Edition Springer (required)
• Introduction to R (see Computing Page)
• R in a Nutshell - Joseph Adler, published by O'Reilly (recommended)
• Matrix Algebra from a Statistician's Perspective - David Harville (1997) Springer-Verlag (or any good text on linear/matrix algebra)
• Statistical Inference - Casella and Berger (or any Mathematical Statistics text for distribution theory)

Purchase books from the bookstore or order directly from Springer, Amazon or other sites.