Seminar Notice
Statistical Laboratory
Iowa State
University
DATE AND TIME: Monday, September 12, 2005,
4:10 p.m.
PLACE: 319
Snedecor
SPEAKER: Richard L. Smith, Department of
Statistics and Operations Research, University of North Carolina, Chapel
Hill
TITLE: Kriging With Estimated Parameters
ABSTRACT
"Kriging" is the
name commonly given to a technique of prediction in spatial processes. In one
very common version of the technique (known as "universal kriging"), the mean of
the process is represented as a linear function of covariates while either the
covariance function or the variogram is represented parametrically. However in
most representations, the estimation of covariance or variogram parameters is
treated independently of the actual prediction, so that resulting prediction
variances or prediction intervals do not take account of the estimation of model
parameters. Bayesian methods do not suffer from this objection, but they are
more computationally intensive and it is still unknown whether Bayesian
prediction intervals have appropriate frequentist coverage
probabilities.
In this talk, I propose a new approach. Asymptotic
expansions are used to examine the true coverage probabilities of prediction
intervals calculated by both frequentist and Bayesian methods, and suggest
substantive grounds for preferring the Bayesian approach. I also explore
applications to the design of monitoring networks. The latter context leads to a
surprising reinterpretation of some recent work by Zhengyuan Zhu and Michael
Stein.
COFFEE: 3:45 p.m., 104 Snedecor
Hall
Seminar schedules and abstracts are available via WWW:
http://www.stat.iastate.edu/
Jeanette La Grange
Department of Statistics
102
Snedecor
Iowa State University
Ames, IA 50010-1210
515 294-3440
(office)
515 294-4040 (fax)
http://www.stat.iastate.edu/directory/staff/jeanette.html