Stanford University, USA
Statistical methods as social technologies versus analytic tools: Implications for metascience and research reform
Statistical methods can be viewed through different lenses; as technical tools to detect signals and quantify uncertainty, or as foundations for making epistemic claims. The use of similar technical tools in different sciences can obscure the profoundly different epistemic cultures across the sciences. Elements of these cultures include what constitutes sufficient grounds for making a claim, e.g. that an intervention “works”, and who is qualified to make that judgment. If these differences are not acknowledged, which they rarely are, debates about how to improve the use of statistics in science often miss the mark, never reaching closure. This will be shown with recent empirical work on whet- her specialized statistical review is needed for research papers, efforts of a major science funder (PCORI) to improve statistical practice in its funded research, and controversy about the use of statistical significance in published research. I will discuss what this means for metascience and for improving statistical practice.