This item has been filed in | Events, Extra Credit, Extra Credit - Medical
Print This Post Print This Post

Division of Biostatistics, Department of Epidemiology and Public Health Seminar

Mar
7
11:30 am

Ralph G. O’Brien, of the Department of Epidemiology and Biostatistics at Case Western Reserve University, will present “Direct Statistical Inference Requiring Only a Touch of Bayes: A ‘March of Science’ Schema for Data Analysis” on Monday, March 7 at 11:30 a.m. in the Clinical Research Building, Gordon Center’s Broad-Bussel Auditorium, first floor. Lunch will be provided.

O’Brien teaches several courses in the Department of Epidemiology and Biostatistics at Case Western Reserve University. Previously, he directed the Collaborative Biostatistics Center at the Cleveland Clinic, was director of biostatistics at the University of Florida, and a professor at the Universities of Tennessee and Virginia. O’Brien’s main contributions in furthering statistical science have been in promoting more effective sample-size analyses for study planning. After his freeware module, UnifyPow.sas, became popular, he collaborated intensively with the SAS Institute in developing PROCs POWER and GLMPOWER. He has given dozens of workshops and tutorials on this topic and has received the American Statistical Association’s Excellence in CE award and the Distinguished Achievement Award from the ASA’s Section on Teaching Statistics in the Health Sciences. He has served the ASA in many ways, including on its board of directors, and is now chair-elect of the ASA’s Section on Statistical Consulting. He is a fellow of the ASA.

Abstract
Maxine Chance, a statistical scientist, collaborates with Will Treatrite, an expert in tropical diseases. They first discuss problems inherent in traditional (frequentist) statistical inference: (1) The point null hypothesis is virtually never true. (2) The two-sided “alternative” hypothesis rarely makes scientific sense. (3) The p-value is only “circumstantial” evidence and is among the most commonly misunderstood concepts in all of science. Chance knows that Bayesian methods offer elegant alternatives, but also sees how their complexities thwart their everyday use. Therefore, she is now proposing a general solution that combines frequentist test statistics with “a touch of Bayes.” First, she illustrates how her method handles data from Treatrite’s series of clinical trials on a malaria-like disease called furlingia (fictitious). Then she uses it to handle data from real studies that received extensive media coverage, including a “preliminary” trial that convinced a major pharmaceutical company to invest $1.3 billion to purchase a patent for a drug believed to reduce atherosclerosis but that had so far shown only a tiny chance of efficacy, and the controversial meta analyses of Nissen and Wolski (2007, 2010), which sparked the FDA’s extensive safety review and its resulting major relabeling of the popular Type 2 diabetes drug Avandia (rosiglitazone).
This presentation is designed to reach a broad audience, and thus will not attempt to cover the mathematical or computing underpinnings of the methods illustrated.

Share this Listing:
  • Twitter
  • Facebook
  • MySpace
  • LinkedIn
  • del.icio.us
  • Google Bookmarks
  • NewsVine
  • Digg

Comments are closed.

  • Related Stories
  • Tags
  • Popular
  • Subscribe
  • Subscribe to the Veritas RSS Feed
    Get updates to all of the latest Veritas posts by clicking the logo at the right.

    You can also subscribe to specific categories by browsing to a particular section on our site and clicking the RSS icon below each section's header.

UM Facebook

UM Twitter