Abstract: This lecture explores the challenges faced by a statistician in a government regulatory environment. With a need to maintain consistency when applying inferential methods to assess regulatory criteria, the statistician is often met with the refrain “but that is the way we have always done it.” This despite the fact the assumptions of the inferential method are not met and often lead to erroneous conclusions. There is a bias toward inferential methods, toward reducing the data to a single test statistic, and ignoring the story hidden in the data.
Six examples that break out of the “but that is the way we have always done it” paradigm are discussed. In two examples, visualization rather than p-values are used to reveal the story and in two others the risks of blindly relying on p-values and not evaluating model assumptions are examined. The final two examples use exploratory data analysis techniques rather than inferential methods to examine the relationship between cigarette physical design parameters and the tobacco, nicotine, and CO2 yields in mainstream smoke and to explore the structure of a complex molecule used in the treatment of muscular dystrophy.
Watch Online: For those who are unable to attend in person, this seminar will be streamed live via WebEx. Simply follow this link and click the "Join" button to join our video feed. Audio can be accessed by calling (855) 749-4750 and entering the meeting code 648 373 386.Event Contact: