Math @ Duke

Publications [#165687] of Benoit Charbonneau
Papers Published
 Juli Atherton, Benoit Charbonneau, Xiaojie Zhou, David Wolfson, Lawrence Joseph and Alain C. Vandal, Bayesian optimal design for changepoint problems,
Canadian Journal of Statistics, vol. 37 no. 4
(2009),
pp. 495513 [cjs.10037]
(last updated on 2009/12/14)
Abstract: We propose, for the first time, optimal design for changepoint problems. Suppose
that a sequence of observations is taken in some subinterval of the real axis. If the
distribution of the sequence changes at some unknown location then we refer to this
location as a changepoint. Changepoint inference usually concerns location testing
for a change and/or estimating the location of the change and the unknown parameters of the distributions before and after any change. In this paper, we investigate
Bayesian optimal designs for changepoint problems. We find robust optimal designs
which allow for arbitrary distributions before and after the change, arbitrary prior
densities on the parameters before and after the change, and any logconcave prior
density on the changepoint. We define a new design measure for Bayesian optimal
design problems as a means of finding the optimal design itself. Our results apply
to any design criterion function concave in the design measure. We show that our
method extends directly to a setting in which there are several paths all with the
same changepoint.


dept@math.duke.edu
ph: 919.660.2800
fax: 919.660.2821
 
Mathematics Department
Duke University, Box 90320
Durham, NC 277080320

