Statistics and Data Science Seminar: "Random-effects models and the statistics of speciation"

Speaker: Peter McCullagh, John D. MacArthur Distinguished Service Professor, Department of Statistics, University of Chicago

Abstract: The talk is concerned with random-effects models and their application in experimental science. I will discuss in detail two specific experiments, both involving animal behavior, plus technical matters as they arise in context. The role of the experimental design in the formulation of stochastic models will be discussed, with emphasis on the baseline and on the mathematical distinction between covariates and relationships. Technical matters that arise include the following: Covariance functions; Compatibility of factorial models with randomization; REML and likelihood-ratio statistics;  Limitations of Wald tests;  Fitted versus predicted values. Remarks on software and computation. The talk is based mainly on a forthcoming book titled "Ten Projects in Applied Statistics", but also partly on joint work with Heather Battey at Imperial College, London.

Host: Debashis Mondal