Bayesian and Nonparametric Statistics - Teaming up two opposing theories for the benefit of prognostic studies in Covid-19

Department of Medical Statistics

The SARS-CoV-2 pandemic has become a global crisis from economical, societal and medical point of views. Identifying risk factors for a severe course of Covid-19 presents major statistical challenges with dramatic consequences for both the treatment (intervention) on a patient level and society as a whole. However, adequate analysis of Covid-19 data for prediction of the disease course poses numerous challenges, as for example small sample sizes, cluster effects, presence of competing risks or heterogeneity. We aim to minimize the risk of false conclusions with innovative methods addressing these challenges. To this end, we will develop and apply meta-analysis methods to judge the models’ predictive performances. As most conventional meta-analyses tools rely on large sample or Gaussian approximations that are highly questionable in most Covid-19 studies, new robust but nevertheless powerful techniques are needed. We therefore follow an approach that combines novel techniques from the two opposing fields of nonparametric and Bayesian statistics. The general idea is to generate robust confidence distributions from parametric and nonparametric resampling methods for each individual study. As a result we can flexibly derive point estimates, tests, p-values and frequentists prediction as well as Bayesian credible intervals to address the challenges. The envisaged approach even allows for the inclusion of expert knowledge by means of choosing proper a priori or additional confidence distributions in the fusion approach.

This project is a joint venture of the Department of Medical Statistics at the University Medical Center Göttingen, the Chair Mathematical Statistics and Applications in Industry at the Technical University of Dortmund and the Institute of Biometry and Clinical Epidemiology at the Charité Universitätsmedizin Berlin. The project is funded by the Volkswagen Foundation within the initiative “Corona Crisis and Beyond - Perspectives for Science, Scholarship and Society”.

Investigators and Collaborators

Professor Tim Friede (principal investigator)

Dr. Christian Röver (collaborator)

NN (PhD student)
Department of Medical Statistics
University Medical Center Göttingen

Professor Frank Konietschke (principal investigator)
Institute Biometry and Clinical Epidemiology
Charité Universitätsmedizin Berlin

Professor Markus Pauly (principal investigator)
Chair Mathematical Statistics and Applications in Industry
Technical University of Dortmund



Univ.-Prof. Dr. Tim Friede

Univ.-Prof. Dr. Tim Friede

contact information


Follow us