Statistics 600: Regression Analysis

Instructors

Kerby Shedden
kshedden@umich.edu
277 West Hall
Office hours: Monday 3-4, Thursday 5-6

Brian Manzo bmanzo@umich.edu
Office hours: Tuesday 4-5:30 (remote), Wednesday 10-11:30 (1720 Chemistry)

Course description

Statistics 600 is an advanced introduction to regression analysis. The course is fast-paced, and focuses on the motivation, construction, and statistical properties of classical and modern regression procedures.

The following topics will be covered: (1) a comprehensive treatment of linear models for independent observations using least squares estimation, with some discussion of non least-squares approaches; (2) regression methods for dependent data, including generalized least squares, estimating equations, and multilevel models; (3) generalized linear models and generalized estimating equations (GEE); (4) alternative approaches to regression, including quantile regression, dimension reduction regression, and smoothing-based methods; (5) issues related to data collection, study design, and interpretation of results including causality, missing data, selection bias, and designing studies.

The class includes a lab that meets each week. The lab will focus on computing, simulation studies, and data analysis skills, and will also be an opportunity to discuss homework sets.

Regular attendance at the lecture and lab is expected.

Prerequisites

A solid background in linear algebra (theory of linear spaces – basic matrix algebra is not sufficient); knowledge of regression at the level of Statistics 500; knowledge of probability and statistical theory at the level of Statistics 510-511 or Biostatistics 601-602; basic programming skills.

Coursework

The final exam is on Monday, December 20th from 1:30-3:30 and will be taken remotely via Gradescope. This is a synchronous exam and you must be available at this time, with a computer and internet connection to take the exam (you do not need to be on campus however). There will also be one in-class exam on November 3rd. Problem sets will be given roughly bi-weekly during the semester. Problem sets will be posted to the course web page, and will be due in class approximately two weeks after being posted. A capstone project covering the whole course will be due on a date in December TBD.

Grading

The final course grade will be weighted 25% from the problem sets, 25% from the midterm, 25% from the final exam, and 25% from the capstone project.