Methodological Research
The Institute of Medical Statistics at the Center for Medical Data Science at the Medical University of Vienna is dedicated to advance statistical methodologies in medical research. Besides methods development, we implement these methods in open source software, aiding in their practical application.
In particular, we focus on the following areas:
Adaptive Designs & Platform Trials
Carrying forward a longstanding tradition of research in adaptive clinical trial design, founded by Peter Bauer, the working group “Adaptive Designs” develops statistical methodology and corresponding software to improve the efficiency of exploratory and confirmatory clinical trials. Building on the research in adaptive designs, we develop statistical methods for the design and analysis of platform trials, a novel trial design that integrates multiple related randomized clinical trials into a single study, handling complexities such as multiplicity adjustment, inclusion of non-concurrent controls, optimal allocation, adaptations, planning and adjusting of sample size, and randomization procedures. Furthermore, we apply adaptive designs to high-dimensional genetic studies, including microarray analyses, RNA-seq, and proteomic studies.
Multiple Hypotheses Testing & Survival Analysis
We develop innovative methods for statistical inference when multiple hypotheses are examined simultaneously. Our primary focus is on their application in randomized clinical trials and high-dimensional studies. Additionally, we design methods for analysing time-to-event data, with an emphasis on between-group comparisons under non-proportional hazards.
Planning and Analysis of Animal Trials & Non-Linear Bayesian Modelling
In relation to animal trials, we evaluate research processes to ensure key quality criteria, such as analysis plan pre-specification and protocol-outcome consistency, are met. This approach is intended to enhance the reproducibility of animal trials. In a separate domain of our work, we develop a flexible non-linear Bayesian modelling framework based on a genetically modified mode jumping Markow chain Monte Carlo (MCMC) algorithm, which is integral for posterior model estimation.In relation to animal trials, we evaluate research processes to ensure key quality criteria, such as analysis plan pre-specification and protocol-outcome consistency, are met. This approach is intended to enhance the reproducibility of animal trials. In a separate domain of our work, we develop a flexible non-linear Bayesian modelling framework based on a genetically modified mode jumping MCMC algorithm, which is integral for posterior model estimation.