B1122
Title: What can statistics do for Open Science and what can Open Science do for statistics?
Authors: Sabine Hoffmann - Ludwig-Maximilians-Universitaet Muenchen (Germany) [presenting]
Abstract: In recent years, the reliability of scientific findings has been investigated through replication and multi-analyst studies, in which multiple teams of researchers are asked to answer the same research question on the same data set. These two types of studies provide increasing evidence that classical statistical methods which only focus on sampling uncertainty convey a disproportionate level of certainty and thereby yield overconfident results, leading to what has been referred to as replication or statistical crisis in science. While these issues have fuelled debate and received considerable attention both in the scientific community and beyond, the involvement of statisticians in finding causes and solutions to this crisis has been surprisingly limited. We will give an overview of ways in which statisticians can contribute to methodological challenges in the Open Science movement and ways in which the Open Science movement can help tackle longstanding methodological challenges. In particular, we will give an overview of ideas on how we can make evidence from different study designs and different analysis strategies comparable, how we can derive uncertainty intervals that account for sampling uncertainty and analytical variability and how we can assess the extent of selective reporting in methodological research.