Statistics When the Model is Wrong

Jeff Miller, Assistant Professor of Biostatistics, and Pierre Etienne Jacob, Assistant Professor of Statistics at FAS, were recently awarded funding by the Radcliffe Institute for a proposed Exploratory Seminar – Statistics when the Model is Wrong.

Miller and Jacob are interested in exploring solutions to model misspecification – the premise that statistical methods rely on model assumptions that inevitably lead to approximation errors, and that these errors can multiply as the size of the data set grows.  

Because of the recent explosion in the data volumes used in research, and because the issue of model misspecification underlies some of the basic problems with reproducibility, the challenge of developing methods robust to model misspecification is currently being tackled across fields ranging from statistics, to economics, to biomedical science.

The goal of Miller and Jacob’s seminar, which will be held during the academic year 2017 – 2018 at the Radcliffe Institute, is to take an interdisciplinary approach to the problem, bringing together a diverse group of researchers to work on scalable inference for robustness to model misspecification, in order to share insights and identify key problems.