Campus Tilburg University hoogleraren

PhD Defense X.M. Kavelaars

Date: Time: 10:00 Location: Aula

A Bayesian multivariate framework for analysis and decision-making with multiple binary outcome variables

  • Location: Cobbenhagen building, Aula
  • Supervisor: Prof. M.C. Kaptein PDEng
  • Co-supervisor: Dr. Ir. J. Mulder

We still offer a live stream for our ceremonies.

Live stream


In this doctoral research project, we focused on analyzing data from Randomized Controlled Trials (RCTs) with two or more outcome measures that have only two possible responses (for example, "yes" and "no"). Although multiple outcome measures are often measured, they are typically analyzed separately. By analyzing them together, we gain a more comprehensive understanding of treatment effects. This approach allows us to assess side effects against effectiveness or simultaneously capture multiple aspects of effectiveness.

Designing RCTs with multiple outcome measures and conducting the corresponding statistical analyses, however, introduces additional complexity compared to analyzing a single outcome measure. We need to consider the interrelationships between the outcome measures in the analysis, provide a clear definition of "better" or "more effective," and use an appropriate method to determine the required number of participants. Since we do not want to expose patients to suboptimal treatments, all of these considerations are aimed at limiting erroneous conclusions. We want to avoid falsely concluding that one treatment is superior to the alternative while also ensuring that effective treatments are not overlooked.

In this dissertation, we propose a framework for the analysis and decision-making process with multiple outcome measures, incorporating this additional complexity. Moreover, the framework has been expanded to identify differences among patients. This helps to personalize treatments and prescribe the right treatments to patients with specific characteristics. and is suitable for analyzing clustered data. Lastly, it has also been extended for the analysis of clustered data.