BIOS 625: Categorical Data Analysis & Generalized Linear Models
Spring 2018


Midterm II:
Prob 1: aspiration dataset
Prob 2: cornfield dataset
Prob 3: crabs dataset


Class Information

Instructor Dipankar Bandyopadhyay
Department of Biostatistics, School of Medicine,
Virginia Commonwealth University
One Capitol Square [OCS], Room # 737
E-mail: dbandyop at vcu dot edu
Teaching Assistants None

Class Schedule Tuesday/Thursday
10:00 AM - 11:50 AM
OCS, # 5009 (Classroom)

Office Hours Thursdays [OCS, # 737] @ 2:00 PM - 3:00 PM, or by appointment

Required Texts A. Agresti: Categorical Data Analysis, 3rd Edition, Wiley, 2013
Errata sheet

Anette J. Dobson & Adrian G. Barnett: An Introduction to Generalized Linear Models,
3rd Edition, CRC Press/Chapman & Hall, 2008

General Class Outline and Reading Blocks

  • Course Outline
  • Introduction (Agresti Chapter 1)
  • Contingency Tables (Agresti Chapters 2, 3)
  • Generalized Linear Models (Agresti Chapter 4)

  • Homework assignments + Other Assignments + class participation (40% total)
  • Two midterm exams (20% each)                                         
  • Final exam (20%)                        

Lecture Notes*

Date Lecture Title
01/16, 01/18 Class Introduction
Lecture 1: Introduction
01/23, 01/25 Lecture 2: Inference on Binomial parameters
Lecture 3: Inference on Multinomial Parameters
Reading Assignment
Lecture 4: Measures of Association and Variance Estimation
Computing: Exercise 1; SAS1
                 Exercise 2; SAS2
                 Program1; Program2
01/30, 02/01 Lecture 5: Contingency table - I
Lecture 6: Contingency table - II
[Following the book:Notes1, Notes2]
02/06, 02/08, 02/13 Lecture 7: Inference for Contingency Tables - I
Lecture 8: Inference for Contingency Tables - II
02/15, 02/20 Testing Independence
Ordinal Measures
On Chi-square partitioning
Midterm samples
Midterm 1 [on 03/01],
Lecture 9: Generalized Linear Models - I
02/27, 03/01 Lecture 10: Generalized Linear Models - II, Generate Random Exponentials
Look at this interesting paper on overdispersion
03/06, 03/08 Spring Break
03/13 Lecture 11: Logistic Regression - I
SAS codes: Crabs, Crabs1, Grouping Example
Reading material: Binomial Link functions
                         A recourse on the cloglog link.
03/15 Midterm-I Solutions
Lecture 12: Logistic Regression -II
Reading material: Separation problem in logistic regression. Ready SAS stuffs
                          Bayesian suggestions [for more advanced readers]
03/20, 03/22,03/27 Lecture 13: Building, Checking and Applying Logistic Regression
The SAS Power procedure
Lasso in 1 page; Original lasso paper
Penalized Likelihood:Intro
Crabs (W+D): SAS code
Aggregate statement
03/29 Lecture 14: Alternative Binary Response Models
More on FIRTH
EXACT logistic regression
SAS Notes: Exact Conditional Logistic Regression, Codes
Bayesian Analysis in SAS: Manual
GAM in SAS; More on GAM
04/03 Lecture 15: Multinomial Responses
SAS Count Data
Proc NLMIXED Intro;;
Note: Midterm 2 [Takehome], April 20 (Friday, 4:30 PM) -- April 24 (Tuesday, 5 PM)
04/05, 04/10 Lecture 16: Loglinear Models; Read Brown's test
Datasets from HW # 5:; marijuana.txt;
Cohen paper [HW # 5]
04/12 Lecture 17: Matched pairs, Generalized McNemar test
04/17, 04/19 Lecture 18: Marginal Models; depress dataset; QIC
SAS documentation for Proc GENMOD
Short Intro in Informative Cluster Size
Goodness-of-fit for the GEE model [Paper1, Paper 2]; SAS code from Paper 1
HW6 ; Efron's Statistical Science paper
04/24, 04/26 Lecture 19: Random Effects Model - I
Bridge Density
Another popular read on Conditional vs Marginal models
A nice primer on Importance Sampling
05/03 No class on 05/01
Lecture 20: Random Effects Model - II
Cumulative link models: Interpretation & other details
SAS codes exploring Zero Inflation, Hurdle Regression and FMM
Extras Bayesian Models for Categorical Data [Agresti's notes]
Download WinBUGS here
Download for Mac/Linux here
Steps to run WinBUGS
Some examples [categorical data]: Seeds, Salmonella, Epilepsy
Final Takehome: May 4 [4:30 PM] - May 9 [5:00 PM]
* Printing instruction: The .pdf files can be printed more than one slide to a page by accessing the advanced printer options on your printers. Check which ones in 
   the Department are capable of doing so.
* Acknowledgements to Dr. Haitao Chu [Univ. of Minnesota] and Dr. Tim Hanson [University of South Carolina]