STAT34700: Generalized Linear Models
Winter 2020
Winter 2020
Instructor:
Jingshu Wang [email protected] Teaching Assistants: Nathan Gill [email protected] Zehao Niu [email protected] Jason Willwerscheid [email protected] |
Class Schedule:
Tuesdays and Thursdays 2:00 PM–3:20 PM in Eckhart 133 Office Hours: Mondays 2-3 pm by Nathan Gill in Jones 304 Wednesdays 1-2 pm by Nathan Gill in Jones 304 Fridays 2-3 pm by Zehao Niu in Jones 304 Tuesdays 3:30-4:30 pm by Jingshu Wang in Jones 317 |
Description
This applied statistics course is a successor of STAT 343 and covers the foundations of generalized linear models (GLM). We will discuss the general linear modeling idea for exponential family data and introduce specifically models for binary, multinomial, count and categorical data, and the challenges in model fitting and inference. We will also discuss approaches that supplement the classical GLM, including quasi-likelihood for over-dispersed data, robust estimation and penalized GLM. The course also covers related topics including mixed effect models for clustered data, the Bayesian approach of GLM and survival analysis. This course will make a balance between practical real data analysis with examples and a deeper understanding of the models with mathematical derivations.
This applied statistics course is a successor of STAT 343 and covers the foundations of generalized linear models (GLM). We will discuss the general linear modeling idea for exponential family data and introduce specifically models for binary, multinomial, count and categorical data, and the challenges in model fitting and inference. We will also discuss approaches that supplement the classical GLM, including quasi-likelihood for over-dispersed data, robust estimation and penalized GLM. The course also covers related topics including mixed effect models for clustered data, the Bayesian approach of GLM and survival analysis. This course will make a balance between practical real data analysis with examples and a deeper understanding of the models with mathematical derivations.
Textbook
Foundations of Linear and Generalized Linear Models by Alan Agresti (required)
Generalized Linear Models by P. McCullagh and J.A. Nelder (optional)
Foundations of Linear and Generalized Linear Models by Alan Agresti (required)
Generalized Linear Models by P. McCullagh and J.A. Nelder (optional)
Syllabus
Week |
Days |
Topics |
Due (Tuesday) |
1 |
T/Th, Jan. 7/9 |
-- |
|
2 |
T/Th, Jan. 14/16 |
HW1 |
|
3 |
T/Th, Jan. 21/23 |
-- |
|
4 |
T/Th, Jan. 28/30 |
HW2 |
|
5 |
T/Th, Feb. 4/6 |
-- |
|
6 |
T/Th, Feb. 11/13 |
HW3 (due Thursday) |
|
7 |
T/Th, Feb. 18/20 |
Mixed effect models (Chap 9) Lecture 12 |
-- |
8 |
T/Th, Feb. 25/27 |
Generalized mixed effect models (Chap 9) Lecture 13 |
HW4 |
9 |
T/Th, Mar. 3/5 |
-- |
|
10 |
T, Mar. 10 (Th - Fri, reading period) |
Review for final |
HW5 |
11 |
Mar. 17 |
Final (1:30 - 3:30 pm) |
-- |
Grading
- Homework assignments: 25%
- There will be 5 assignments in total.
- Late homework will not be accepted for grading (medical emergencies excepted with proof).
- Homework will be submitted through Canvas and is due at 12pm the due date.
- Midterm: 35%
- Final exam: 40%
Supplementary reading materials
Supplementary reading materials
- Website for Agresti's GLM book
- Efron's GLM notes: exponential family Part 1, Part 2, GLM Part 3
- Brown, Cai and Dasgupta's definitive treatment of Interval estimation for a binomial proportion
- Kosuke Imai's slides on Discrete Choice Model.
- Lecture notes on survival analysis from Ronghui Xu at UCSD. We focus the main ideas from her Lecture 1-5.