STA 216 Generalized Linear Models

Instructor: Gary Rosner, 223A Old Chem, grosner@stat.duke.edu

Meeting Times: Monday & Wednesday, 2:20 - 3:35, Room 025 Old Chem
 
Required Text: Applying Generalized Linear Models 

by James K. Lindsey, 

New York: Springer-Verlag New York, Inc., 1997
Supplementary Text: Generalized Linear Models, second edition

by P. McCullagh and J. A. Nelder,

London: Chapman and Hall, 1989

Software:
 
Splus S-Plus 4.5 for Windows, Student Edition- NEW!
by Math Soft, Inc. ©1999 
ISBN: 0-534-35873-X

 

Course Objectives

The overall objective of this course is to expand the notion of regression to a larger class of data models than encountered in linear regression. The expanded regression models fall under the general heading "Generalized Linear Models."

Computing Information

Course Topics

 
Generalized linear models (GLMs): 
  • terminology, 
  • model fitting, 
  • exponential family of distributions 
Linear regression (review):
  • least squares (ordinary and weighted), 
  • maximum likelihood and posterior inference, 
  • residual analysis and model selection.
Loglinear models for categorical data 
Logistic regression 
Polychotomous regression 
Poisson regression 
Failure-time data (survival analysis) 
Extensions (as time permits) to include
  • Growth curves and dynamic linear models 
  • Repeated categorical data 
  • Frailty models 
  • Quasilikelihood and generalized estimating equations (GEEs)

 
 

Course Schedule
 
 
Week Topic
Assignment
Data Sets
Due
Notes
30 Aug Introduction, terminology, 
review of exponential families
6 Sept Review of linear regression
 hw1
 Timber.dat
13 Sept
 Sept6.ps
13 Sept Logistic regression
22 Sept
 hw2
  GHQ.dat ,  cholesterol.dat
29 Sept
27 Sept Multinomial/Poisson models
4 Oct
13 Oct
18 Oct
 hw3
20 Oct
25 Oct BUGS with Dr. Nicky Best  BUGS manual;
Examples vol. 1 ; CODA manual;
CODA ver 4 addendum		  homework.ps
 hb.dat
 hbwin.dat
 suicide.dat
 hospital.dat		 The following will get attach the CODA functions to your S-plus session at Duke Statistics.
attach("/usr/local/lib/CODA/.Data")

Please use file "hbwin.dat" for WinBUGS and "hb.dat" for Bugs.
8 Nov More WinBUGS & Survival Analysis  Survival analysis example data
10 Nov Midterm (due 17 Nov at start of class)  Midterm exam