31st International Summer School of Epidemiology

 

COURSE 1: ANALYTIC METHODS FOR LONGITUDINAL DATA INCLUDING TIME-VARYING COVARIATES

Steve Marshall

Longitudinal Data Analysis is an important topic in epidemiology. We often seek to gain insight into causal relationships by prospectively following subjects over time. However, it is important to recognize that exposures and confounders are often changing over time. Thus, it is important to understand analytic methods for longitudinal data in order to isolate effects of specific exposures while controlling for time-dependent covariates. The longitudinal analysis methods taught in this course include generalized estimating equations, generalized linear mixed models with random intercepts and random slopes and survival methods for time-varying exposures. This course will also address pragmatic topics in contemporary epidemiology: Principles of causal inference, estimation versus hypothesis testing and assessing effect modification. This course is intended for those who are proficient with traditional epidemiologic methods and wish to learn advanced methods. The course will be taught using many worked examples in SAS and R. Real data will be provided to allow students to practice different models. Students are also encouraged to bring their own data to the class and interact with the class instructor.