Estimation of treatment effects in clinical trials in the presence of (partial) clustering and trends over time - implementation of mixed models in psycho-cardiology (randomized multicenter trial) and health services research (stepped wedge design)

The research project focuses on how clustered data in the statistical model can best be implemented both in the cohort stepped wedge and the partially nested designs. The key issues to be tackled in model fitting are convergence of estimation algorithms and unbiased estimation, specifically in the presence of both small number of clusters and small cluster size. For the cohort stepped wedge design, a specific use of contrasts is evaluated. Based on an example study with partially nested design, a simulation study is conducted to investigate how a robust model can be implemented in practice.