Prof. Dr. Anne-Laure Boulesteix

We are happy to announce that the Reinhart Koselleck grant application with the title

"Design, interpretation and reporting of empirical evaluations of statistiscal methods"

at the interface between statistics/biometrics and metasciencewas accepted for funding by the DFG (German Research Foundation).

Abstract: Methodological statistical research, i.e. the development and investigation of statistical methods, is not immune to the problems that are well known to affect the reliability of empirical studies in other scientific fields such as medicine. These issues, which can lead to over-optimistic conclusions and difficulties in translating research results into practice, include poor study design and selective reporting. In this context, the overarching aim of the proposed project is to strengthen the validity and utility of methodological research and literature by improving the methodology of comparisons of statistical methods. The focus will be on the design, interpretation, and reporting of experiments investigating methods, including both statistical simulations and real-data-based benchmark studies. The improvement of methodological research methodology should in the long run also lead to an increase in research quality in fields of empirical research that apply these methods, such as medicine. More precisely, we aim to (i) gather evidence on the roots of the problems and (ii) demonstrate and evaluate potential solutions in order to improve the quality of methodological studies, i.e. to make methodological studies less biased, to make their results more precise and nuanced, to improve methodological research practice towards a better translation of methodological results into applications, and to maximize the gained information on the behavior and performance of methods given a fixed computational budget. These objectives are addressed through a meta-scientific approach using various methodologies, including literature reviews, case studies, simulation studies, and Delphi surveys.