The Bayesian Statistical Science Section of the ASA provides a forum for statisticians and people who have interest in the Bayesian paradigm. The broad objectives of the Section are: to encourage research on theory and methods of statistical inference and decisionmaking associated with Bayes' theorem and to encourage the application and proper use of Bayesian procedures in the behavioral, biological, managerial, engineering, environmental, legal, medical, pharmaceutical, physical, and social sciences.
Upcoming Web-Based Lectures
Title: NIH Opportunities for Theory, Modeling, Computation, and Statistics Researchers
Presenter: Gregory Farber, Director, Office of Technology Development and Coordination, National Institute of Mental Health
Date and Time: Wednesday, June 17, 2015, 1:00 p.m. - 2:00 p.m. Eastern time
Sponsor: Mental Health Statistics Section
Registration Deadline: Monday, June 15, at 12:00 p.m. Eastern time
The National Institutes of Health works to improve health and save lives. Quantitative researchers play a key part in these efforts. This talk will discuss the various programs that might be of interest to the statistics community. Areas that will be covered include the Big Data to Knowledge (BD2K) effort, the BRAIN Initiative, new NIMH expectations for data sharing, and some of the many databases that make data freely available to the research community.
Title: Propensity Score Methods for Estimating Causal Effects in Pharmaceutical Research
Presenter: Elizabeth Stuart, Johns Hopkins Bloomberg School of Public Health
Date and Time: Wednesday, June 24, 2015, 12:00 p.m. - 2:00 p.m. Eastern time
Sponsor: Biopharmaceutical Section
Registration Deadline: Monday, June 22, at 12:00 p.m. Eastern time
Propensity scores are an increasingly common tool for estimating the effects of interventions in observational ("non-experimental") settings and for answering complex questions in randomized controlled trials. They can be of great use in pharmaceutical and health services research, for example helping assess broad population effects of drugs, devices, or biologics already on the market, especially investigating post-marketing safety outcomes, or for answering questions regarding the outcomes of long-term use using claims data. This webinar will discuss the importance of the careful design of observational studies, and the role of propensity scores in that design, with the main goal of providing practical guidance on the use of propensity scores to estimate causal effects. The webinar will briefly cover the primary ways of using propensity scores to adjust for confounders when estimating the effect of a particular "cause" or "intervention," including weighting, subclassification, and matching. Topics covered will include how to specify and estimate the propensity score model, selecting covariates to include in the model, diagnostics, and common challenges and solutions. Software for implementing analyses using propensity scores will also be briefly discussed. The webinar will also highlight recent advances in the propensity score literature, with a focus on topics particularly relevant for pharmaceutical contexts, including prognostic scores, covariate balancing propensity scores, methods for non-binary treatments (such as dosage levels of a drug or when comparing multiple drugs, devices, or biologics simultaneously), and approaches to be used when there are large numbers of covariates available (as in claims data).
Title: Mixed Models for Longitudinal Categorical Outcomes
Presenter: Donald Hedeker
Date and Time: Thursday, July 16, 2015, 12:00 p.m. - 2:00 p.m. Eastern time
Sponsor: Mental Health Statistics Section
Registration Deadline: Tuesday, July 14, at 12:00 p.m. Eastern time
This webinar will focus on analysis of longitudinal data using mixed models. Because dichotomous, ordinal and nominal outcomes are common in many areas of research, this webinar will mainly consider the mixed logistic regression model, and generalizations of it. Specifically, the following models will be described: mixed logistic regression for dichotomous outcomes, mixed logistic regression for nominal outcomes, and mixed proportional odds and non-proportional odds models for ordinal outcomes. The latter models are useful because the proportional odds assumption of equal covariate effects across the cumulative logits of the model is often unreasonable. Use of the program SuperMix for these models will be described and illustrated.
*** More information and registration at the webinar page.