Beyond Multiple Linear Regressions with R

This course is primarily intended for Research Master students. There is only a limited number of spots available for PhD students. NB. CPE-students need to register via the CPE-office.

Course description
Traditional introductory statistics courses often rely on models that fit data under a set of strict assumptions: continuous response variables, independent observations, and residuals with constant variation. However, these assumptions are frequently violated in practical applications, presenting a challenge for standard statistical methods.

In this course, "Beyond Linear Regression", we delve into more advanced modeling techniques to address these challenges, with a particular focus on mixed effects models. We will begin by revisiting linear regression models, placing an emphasis on model building and coding aspects, such as effects coding versus dummy coding. Understanding these concepts serves as a crucial intermediate step towards moving from fixed effects to random effects models.

Mixed effects models are versatile tools that allow for the inclusion of both fixed and random effects, accommodating data structures such as clustered or longitudinal data, where observations are interdependent.

Finally, we will move beyond the analysis of continuous outcomes by exploring how random effects can be incorporated into other types of regression models, such as logistic regression models for the analysis of binary response variables.

Learning objectives
By the end of this course, you will be equipped to:

• Understand and implement linear regression models with a focus on model building
• Grasp the differences between effects coding and dummy coding, and their implications for modeling
• Apply generalized linear models to binary and other types of response variables
• Apply mixed effects models to handle interdependent observations and complex data structures
• Effectively communicate results from statistical analyses to a general audience

Approach
The course will extend over a five-week period, in which the first four weeks will be dedicated to the learning and practicing of the material in selected chapters from the two textbooks, and the fifth week will be dedicated to the final assignment.

Textbooks
Both books are freely available online. Print copies are also available for purchase.

Roback, L. Beyond Multiple Linear Regression, CRC Press, 1st edition; 2021: https://bookdown.org/roback/bookdown-BeyondMLR/

Leyland AH, Groenewegen PP. Multilevel Modelling for Public Health and Health Services Research: Health in Context. Cham (CH): Springer; 2020:
https://link.springer.com/book/10.1007/978-3-030-34801-4

Prerequisites
In addition to an introduction course in statistics, you will need to feel comfortable using R, which has become the lingua franca of data science in many academic circles.

Exam / attendance
Exam is mandatory. 80% attendance is required.