Analysis of Repeated Measurements 2023
OrganisatorDr. R. Tsonaka
The course covers statistical methods to be used in the situation where one or more outcome variables are repeatedly measured in time on the same experimental unit. For instance, in a clinical trial, the outcome variable can be measured at baseline and at different times during the treatment period. For this type of data, traditional regression models cannot be used since outcomes of the same subject may be correlated, and this should be taken into account in the statistical model. Other examples of studies where there is a similar dependency in the outcome measurements are cluster randomised trials and meta-analysis. In the last one or two decades, much progress has been made in the development of new methods of analysis for dependent outcome data. In recent years several of these new methods have been implemented in commercially available computer packages. In the course, first an overview of classical approaches to repeated measurements will be given. Then modern methods are introduced. For approximately normally distributed response, we will discuss the Linear Regression model with correlated errors and the Linear Mixed Model. For non-normal response, the generalized estimating equations (GEE) approach for marginal models and the Generalized Linear Mixed Models (e.g., random effects logistic regression) are discussed. Examples of clinical and epidemiological applications will be given. In the computer labs, exercises with R and SPSS are used to get experience with applying these new methods to real data.