ORIE 671 -- Intermediate Applied Statistics -- FALL 2003 TIME: MWF 1:25 -- 2:15pm, OH 218 PREREQUISITES: Linear Algebra and Matrices + ORIE 670 or equivalent INSTRUCTOR: Jim Booth Visiting Professor Operations Research and Industrial Engineering 216 Rhodes Hall Phone: 255-9137 OFFICE HOURS: MW 2:30 -- 3:30pm, Th 10 -- 11am SUMMARY: This course will focus on the theory and application of linear and generalized linear models and related statistical topics. Generalized linear models are a very broad family of statistical models, loosely described as follows. The response variable has a distribution in an exponential dispersion family and the mean response is related to covariates through a link function and a linear predictor. GLMs allow a unified theory for many of the models used in statistical practice, including normal theory regression and ANOVA models, loglinear models, logit and probit models for binary data and models for gamma responses and survival data. The statistical package GLIM was specifically designed to fit generalized linear models using the iteratively reweighted least squares algorithm. However, similar model fitting capabilities are now standard in other statistical packages such as R/Splus and SAS. This course will emphasize the use of R for statistical computing and model fitting. TENTATIVE OUTLINE: Outline of generalized linear models Normal theory linear models Binary data Models for polytomous responses Log-linear models Models for continuous responses Extensions to dependent responses Generalized estimating equations (GEE) COURSE TEXT: Generalized Linear Models, 2nd Edition, 1989 by Peter McCullagh and John Nelder: Chapman and Hall Comment: While I will generally follow the content of the text, there will be a considerable amount of material taken from other sources. OTHER REFERENCES: Other sources for course material include: The Theory of Exponential Dispersion Models and Analysis of Deviance by Bent Jorgensen Statistical Modelling in GLIM by M. Aitkin, D. Anderson, B. Francis and J. Hinde: Oxford Science Publications Generalized Additive Models by Trevor Hastie and Robert Tibshirani: Chapman and Hall Multivariate Statistical Modelling Based on Generalized Linear Models by Ludwig Fahrmeir and Gerhard Tutz: Springer Introductory Statistics with R by Peter Dalgaard: Springer GRADING POLICY: There will be two exams (a midterm and a final). Homework will also be assigned periodically and graded. Homework assignments may involve data analysis and computation requiring students to use the R package. In addition, each student will be required to give a presentation on a topic related to the course material but not covered class. I will provide a list of potential topics, but you may also choose your own. Each student's topic must be approved in advance by me. POINTS: Midterm 20, Final 40, Homework 30, Presentation 10.