mathematics colloquium “ nelder memorial lecture”

1
Mathematics Colloquium “Nelder Memorial Lecture” Professor Peter McCullagh FRS (University of Chicago) on: “Some issues connected with likelihood and REML Thursday, March 8 th 2012, 5-6pm LT 308 Computing Department, Huxley Building Reception afterwards in the Huxley Common Room , Level 5, Huxley Building. Abstract: Residual likelihood and residual maximum likelihood estimation have been found to be very useful for variance- components estimation in linear models, particularly for models having more than one variance component. The talk is concerned with marginal likelihood in general and with REML for the computation of likelihood-ratio statistics. Can REML be used for testing the significance of a treatment effect? Can REML be combined with the Box-Cox technique for testing the appropriateness of various transformations?

Upload: linda-brady

Post on 02-Jan-2016

17 views

Category:

Documents


0 download

DESCRIPTION

Mathematics Colloquium “ Nelder Memorial Lecture”. Professor Peter McCullagh FRS (University of Chicago) on: “ Some issues connected with likelihood and REML ” Thursday, March 8 th 2012, 5-6pm LT 308 Computing Department, Huxley Building. Abstract: - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Mathematics Colloquium “ Nelder  Memorial Lecture”

Mathematics Colloquium“Nelder Memorial Lecture”

Professor Peter McCullagh FRS (University of Chicago)on: “Some issues connected with likelihood and REML”

Thursday, March 8th 2012, 5-6pm LT 308 Computing Department, Huxley Building

Reception afterwards in the Huxley Common Room , Level 5, Huxley Building.

Abstract:

Residual likelihood and residual maximum likelihood estimationhave been found to be very useful for variance-components estimation in linear models, particularly for models having more than one variance component.The talk is concerned with marginal likelihood in general and with REML for the computation of likelihood-ratio statistics.Can REML be used for testing the significance of a treatment effect?Can REML be combined with the Box-Cox technique for testingthe appropriateness of various transformations?