Recent Developments in Biostatistics
Chair: Xikui Wang
Organizer: Xikui Wang
Sponsor: Biostatistics Section
[Monday, June 12, 2017 13:30-15:00]
13:30-14:00
Peter X. Song (University of Michigan), Ling Zhou (University of Michigan)
Confidence Estimating Functions for Data Integration
The theory of statistical inference along with the strategy of divide-and-combine for large-scale data analysis has recently attracted considerable interest due to the popularity of the MapReduce scheme. The key to the development of statistical inference lies in the method of combining results yielded from separately mapped data batches. We consider a general inferential methodology based on estimating functions, which allows us to perform regression analyses of massive complex data via the MapReduce scheme, such as longitudinal data, survival data and quantile regression, which cannot be done using the maximum likelihood method. The proposed statistical inference inherits many key large-sample properties of estimating functions. Also we show that the proposed method is closely connected to the generalized method of moments (GMM). Our method provides a unified framework for many kinds of statistical models and data types, which is illustrated via numerical examples in both simulation studies and real-world data analyses.
14:00-14:30
Mikelis Guntars Bickis (University of Saskatchewan), Naeima Ashleik (University of Saskatchewan), Juxin Liu (University of Saskatchewan)
Inference from Survival Data using Imprecise Probabilities
Imprecise probability (IP) is a generalization de Finetti's approach to probability, formalized by Williams (1975) and extensively discussed in a monograph by Walley (1991), who presented these ideas in a JRSS discussion paper (1996). The methodology can be interpreted as Bayesian sensitivity analysis using a set of priors. In his 1996 presentation, Walley introduced a practical form for this inference on discrete data by using a family of Dirichlet priors with a fixed concentration parameter. He proposed that it could also be used for survival data by discretization. Coolen (1997) extended Walley's model to allow for censoring, and Coolen and Yan (2004) proposed a nonparametric predictive inference paradigm which also led to IP conclusions. Bickis (2009) used IP concepts to estimate the hazard function. After reviewing the methodology, we will be presenting an IP version of the log-rank test, which will produce a sequence of imprecise posteriors updated at the time of each death.
14:30-15:00
J. Jack Lee (University of Texas MD Anderson Cancer Center)
Bayesian Adaptive Designs for Efficient Drug Development and Precision Oncology
Clinical trial is a prescribed learning process. Bayesian methods take the “learn as we go” approach and are uniquely suitable for such learning. In recent years, rapid advancements in medicine demand innovative methods to identify better therapies and the most appropriate population in a timely and efficient way. I will first illustrate the concept of Bayesian update and Bayesian inference, then, give an overview of Bayesian adaptive designs in dose finding, predictive probability, multi-arm platform design, and outcome adaptive randomization, etc. Applications including BATTLE trials in lung cancer and I-SPY trials in breast cancer will be given. Bayesian adaptive designs increase the study efficiency, allow more flexible trial conduct, and treat more patients with more effective treatments in the trial but also possess desirable frequentist properties. Perspectives will be given on future development in trial design, conduct, and evaluation to streamline and speed up drug approval.