STAT 244: Linear and Generalized Linear Models

---

Course goals:

This course presents the theory and application of linear and generalized linear models. Topics include ordinary linear models that usually assume a normally distributed response variable, models for binary and multinomial response data, models for count data, quasi-likelihood and compound models for overdispersed data, and (time-permitting) an introduction to generalized linear mixed models. The class of generalized linear models contains many models commonly used in statistical practice. 

This is not a data analysis class - the focus of the course is on the theory of linear and generalized linear modeling.  If you are seeking practical experience with linear and generalized linear models, you should consider instead Stat 139 and Stat 149, respectively.

Course format:

Lectures (1.25 hours) twice per week, and one discussion section per week.  Lecture time will be spent with the instructor lecturing from prepared notes, with plenty of opportunities for students to comment and ask questions.  The class sections will review the more difficult lecture material, and go through practice problems with heavy student involvement.

Typical enrollees:

This course is aimed primarily at PhD and AM students in Statistics and at upper-level undergraduate statistics students.  In the most recent version of the course (Fall 2022), most of the students were upper-level undergraduate students.

When is course typically offered?

Fall semester.

What can students expect from you as an instructor?

Here's an example comment on my Stat 244 teaching from the Fall 2022 evaluations:

"Mark is an extremely helpful and effective lecturer! You can tell he really cares about the content and about teaching. He is very supportive and lectures are enjoyable. Somehow he makes the material not dry."

I regard myself as communicative and direct, and also respond to email very promptly.

Assignments and grading:

The course will include six bi-weekly homeworks.  In addition, the course will have two take-home exams; one before Spring break, and the second during final exam period.  The overall course grade will be determined in the following percentages:

Homework assignments:  40%

First Exam:  30%

Second Exam:  30%

Sample reading list:

Agresti A (2015). Foundations of Linear and Generalized Linear Models. Wiley. ISBN-13: 978-1118730034. ISBN-10: 1118730038. (Required textbook)

McCullagh P, and Nelder JA (1989). Generalized Linear Models (2nd ed). Chapman and Hall/CRC.  ISBN-13: 978-0412317606. ISBN-10: 0412317605. (Optional reference book)

Enrollment cap, selection process, notification:

No enrollment cap.

Past syllabus:

Last year's syllabus is here.  The main planned differences are (1) there will be no course project, (2) all six problem sets will count towards the final grade, and (3) the exams will be take-home (responding to some concerns about the pressures of in-class exams from last time). 

Absence and late work policies:

The official course policy is no late homeworks. If extenuating circumstances make it difficult to submit a homework on time, a student may petition for a short extension.  I like to release homework solutions promptly after the homework due date, so any extension is likely to be at most a day or two.

Attendance will not be taken at lecture, so absence from lectures is permitted.  You are responsible for learning the course material if you are not able to attend a lecture.  Depending on the room assigned, we will video-record lectures, so that may help for students unable (or unwilling?) to attend lectures.