Sander Greenland, M.A., M.S., Dr.P.H., C.Stat
Department of Epidemiology and Department of Statistics
University of California, Los Angeles
Statistical methods play a pivotal role in health and medical sciences, but not always an enlightened one. Problems well known to academics are frequently overlooked in crucial nonacademic venues such as regulation and litigation, even though those venues can have profound impacts on population health and medical practice. There are many examples of highly credentialed statistical experts using arguments that conceal high prior mass (spikes or point priors) on null hypotheses, including conflation of "nonsignificance" with support for the null and inappropriate use of multiple-testing adjustments. These examples illustrate a more general and universal null bias in standard teaching and practice of statistical hypothesis testing. This bias is not found in the foundational writings of Neyman and Pearson. It may have originated in the transformation of defensible parsimony heuristics into dubious parsimony metaphysics, and confusion of decisions with beliefs, bolstered by fallacious statistical and biological arguments. The "evidence inflation" it produces adds to the burden of cognitive problems in both informal and formal evidence assessments.
All statistics students are expected to attend.