Department of Statistics, The Ohio State University
Statistics and Biostatistics Colloquium Series

In applied statistics, p-values are commonly interpreted as indicating the strength of the evidence against the null hypothesis, despite several cogent and well-established arguments against this practice. The persistent use of p-values in this context underscores both the acute need for measures of statistical evidence in many areas of application, as well as the perceived absence of a viable alternative to the p-value for this purpose. In this talk, I'll begin by considering fundamental properties of evidence measures, drawing heavily on analogy with the measurement of temperature. Next, I'll describe an approach to measurement of evidence based on posterior probability, and illustrate some practical implications of choosing a measure of evidence in connection with gene-mapping in human genetics. Finally, I'll indicate related, ongoing research in the newly-established Center for Quantitative and Computational Biology at Children's Research Institute, in the hope of sparking interest in new collaborations.

Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.