University of Newcastle upon Tyne

School of Mathematics and Statistics

Statistics Seminars 2004-2005

22 April 2005, L401, 2:15pm

Dr Merrilee Hurn, Statistics Group, University of Bath

Abstract

Several years ago, Havard Rue showed how visually improved estimates of binary images could be obtained by using loss functions which were better suited to the particular estimation problem than the usual choices. Can we use similar ideas in other applications, tailoring the loss function to the problem, particularly when the problem is awkward in some way? For example, in a Bayesian mixture problem, how do we estimate the parameters of each component when there is an identifiability issue? This talk will discuss some such problems and the loss functions proposed, including a recent application attempting to estimate an unknown number of QTL (quantitative trait loci) on several chromosomes.

(This work has been joint with Christian Robert, Giles Celeux, Ana Justel and Scott Sisson.)

Return to timetable for 2004-05 seminars

School of Mathematics and Statistics Homepage