SSC Gold Medalist 2022

David Stephens
SSC Gold Medalist
2022

The Gold Medal is awarded to a person who has made outstanding contributions to statistics or probability, either to mathematical developments or in applied work. The award is normally awarded to someone still active in research. The recipient should be Canadian or a permanent resident of Canada, and must have made high quality research contributions to the statistical sciences in Canada. A recipient of the Gold Medal must be a member of the SSC.
 

Dave was born in 1965 in Hereford, England, where he grew up with his parents and four siblings. Interested in mathematics, he registered at the University of Nottingham, mainly because his older brother had studied there. As an undergraduate (BSc, 1986), Dave was fortunate to have Adrian F.M. Smith as a professor and later as his PhD advisor. His thesis was on Bayesian edge-detection in image processing (PhD, 1990).

Between 1990 and 1995, Dave was a postdoctoral fellow and research associate both at Nottingham and at Imperial College London, including 2½ years working on a Ciba-Geigy software development project for population pharmacokinetics using Markov Chain Monte Carlo (MCMC). From 1995 to 2006, he was successively Lecturer and Senior Lecturer at Imperial. He developed interests in statistical genetics and time series analysis, and collaborated, among others, with the British statistician and geneticist Cedric A.B. Smith.

Some of Dave’s notable early contributions include pioneering the use of the Gibbs sampler in change-point identification and developing, with his friend Petros Dellaportas, numerical techniques for Bayesian inference for errors-in-variables models. His talent was further expressed in highly innovative contributions to population-based reversible jump MCMC and the analysis of quantitative locus data where he championed the use of similar methods. His most cited work is on the label switching problem in mixture models.

Dave says that his luckiest break was to have met Erica Moodie at the wedding of his friend Jon Wakefield (Professor of Statistics, University of Washington) in St. John’s, Newfoundland, in 2002. After their own marriage, Dave and Erica both found employment at McGill University in 2006, where Dave was hired as a Full Professor in the Department of Mathematics and Statistics, where he held a James McGill Chair (2011-18), was Department Chair (2015-19) and is now Vice-Dean of the Faculty of Science (2019- ).

Since coming to Canada, Dave has continued to publish his research in the most exalted journals in the profession. His current publication count hovers around 150; approximately two-thirds of his papers have appeared in statistical methodology journals. Dave has also been an incredibly active and caring mentor. At McGill alone, he supervised 10 postdoctoral fellows, and over 40 MSc or PhD students – roughly in equal proportions – graduated under his tutelage; many more are currently working with him. Moreover, he held leading roles as committee Chair (2011-12) for Statistical Sciences at NSERC and as Editor-in-Chief of The Canadian Journal of Statistics (2013-15). He became a Fellow of the American Statistical Association in 2019.

By now, Dave’s contributions are so diverse that it is difficult to summarize them, though they all have a strong Bayesian computational flavor. Among other things, he developed Monte Carlo algorithms for stochastic volatility models driven by Lévy processes, and he designed the first comprehensive and truly Bayesian approach to estimation of marginal structural models by establishing a link between inverse probability weighting and importance sampling. In his many joint papers with Erica, Dave has investigated, among others, properties of ensemble methods for estimating propensity score models, and he has proposed model and variable selection procedures designed for the causal inference setting.

In genetics, Dave made pioneering contributions to Bayesian co-clustering of gene expression profiles. In phylogenetic modeling, he developed original and highly relevant clustering algorithms based on HIV-1 sequence data to identify transmission clusters. In microbiome data analysis, he proposed robust identification of differentially abundant microorganisms between health conditions or treatment groups by constructing networks to estimate the level of co-occurrence between taxa.

Dave Stephens is not one to seek the limelight, but his work shines, and widely so, including outside academic circles. For example, he was approached in 2013 by the World Anti-Doping Agency to help establish and validate, with McGill colleagues, limits for a human growth hormone used to detect illegal additions of this substance. The statistical methods they developed resulted in limits published in Growth Hormone & IGF Research in 2014. This work has been, quite literally, “game changing”: this legally-upheld standard was first implemented in the Sochi Winter Olympics.

In his spare time, Dave enjoys spending time with his family and following sport on TV (mainly soccer, cricket, rugby) or when accompanying his sons Gordie and Jamie to soccer practices and games.


 

The citation for the award reads: 

“To David A. Stephens, for extraordinary research in Bayesian computation and statistical theory, genetics, and causal inference, including applications to Markov Chain Monte Carlo and changepoint detection methods; for extensive editorial work and administrative leadership; for outstanding teaching and mentoring within the statistical community.“

This text was written by Christian Genest and Alexandra M. Schmidt, who made the nomination.