Dr. Tanzy Love is a Bayesian Statistician and an Associate Professor of Biostatistics and Computational Biology at the University of Rochester. She has also held visiting positions at Carnegie Mellon University and Stanford University. Dr. Love has published over 100 papers in refereed books and journals and has served as the primary dissertation adviser for eight PhD students. Dr. Love is currently President of the Rochester Chapter of the American Statistical Association. She is also past President of The Classification Society, an international and interdisciplinary organization devoted to the scientific study of classification and clustering, and a member of the International Federation of Classification Societies.  Dr. Love has been teaching statistics for 25 years and enjoys finding new ways to communicate these ideas. During her spare time, Tanzy is an acrobat who enjoys aerial arts and has been featured in past Bayesian cabaret performances.

Plenary Talk:   An Approach to Simultaneous Model-Based Clustering of Mixed-Type Data with Variable Selection

Abstract:  Motivated by studying diet patterns in the Seychelles Child Development Study, we require a method to discover subpopulations when dietary measures are both quantity (grams of meat consumed) and categories (dairy/no dairy). Our new framework for model-based clustering on data with continuous and discrete variables extends the cluster variance structure framework for model-based clustering of continuous variables with mixtures of Gaussian distributions set forth by Fraley and Raftery (1999). In modeling how each variable contributes to cluster determination, we allow for relationships within and between the continuous and discrete variables. We also modify and extend the likelihood-based variable selection procedures of Raftery & Dean (2006) to accommodate data with variables of mixed-distributional forms. The likelihood contributions of discrete and continuous variables are not comparable. The simulation study results showed that our method is accurate at clustering and variable selection when applied to data with variables of mixed-distributional forms. Applying our clustering and variable selection methods, dietary data shows the diet patterns of pregnant mothers and school-age children. Applying the method to prostate cancer data shows subgroups with different responses to treatment.

Plenary Talk:   Group Sequential Trial Design using Stepwise Monte Carlo for Increased Flexibility and Robustness

Abstract:  Bayesian adaptive clinical trials are becoming increasingly complex, incorporating numerous parameters and degrees of freedom. Optimal analytic approaches for these intricate trial designs are often unavailable, necessitating extensive simulation to control the Type I error (false positive) rate, maximize trial power, minimize trial sample size, and ensure other favorable operating characteristics. This paper proposes a general method to reduce the number of parameters using group stepwise methods and Monte Carlo simulations, significantly decreasing the number of iterations required to identify near-optimal parameters. The key idea is using the Hwang-Shih-DeCani family of error-spending functions, which use just two parameters (an alpha-spending parameter and a beta-spending parameter ) that determine sensible stopping boundaries for efficacy and futility, respectively.  The algorithm then optimally determines stopping boundaries in such a way that power is maximized and overall Type I error is strictly controlled at a predetermined level.  Our method extends classical group sequential designs but does not rely on normality assumptions and can accommodate complex trial designs.  We illustrate in the case of a multi-arm clinical trial with one control arm and k treatment arms, where interim analyses are performed when 30%, 50%, and 70% of the total sample size have been accumulated.  Our approach delivers boundaries that offer significant average sample size reductions under both the null and alternative hypotheses.  Time permitting, we will also briefly discuss the use of the new PhaseV adaptive platform to find optimal designs using our approach.

Gary Sharp is a Professor in the Statistics Department at Nelson Mandela University, a department he has had the honour to serve since 1997. During his time at Mandela University, he has served in many roles, including a seven-year tenure as Department Head. Gary’s research includes a wide range of topics, although his passion lies with analytics in sports, an area for which he is well known. He has supervised several senior post-graduate degrees, including several doctoral graduates. He is an active member of the South African Statistical Association (SASA) and has been instrumental in having the Mandela University Department of Statistics host annual SASA conferences. Gary is a Fellow and past President of SASA, a C2 NRF-rated researcher who has published nationally and internationally. 

Plenary Talk: An introduction to the statistical methods used in selective sporting activities

Abstract: Sporting activities have long since relied on statistical methods to assess and compare teams and/or individual performances. The home team advantage is well documented, and theories abound to support the empirical evidence that when playing a home game, a team is more likely to succeed. Football analytics include using discrete distributions to model a team’s weaknesses and strengths, whilst analysis in cricket competitions has evolved as the format of the game has adapted to commercial interest. This presentation will introduce the audience to the background in sporting analytics and progress to interesting results undertaken by South African statisticians interested in sporting analytics.