N.L. Hjort
http://repub.eur.nl/ppl/8363/
List of Publicationsenhttp://repub.eur.nl/logo.jpg
http://repub.eur.nl/
RePub, Erasmus University RepositoryConstancy of distributions: asymptotic efficiency of certain nonparametric tests of constancy
http://repub.eur.nl/pub/548/
Fri, 20 Sep 2002 00:00:01 GMT<div>A.J. Koning</div><div>N.L. Hjort</div>
In this paper we study stochastic processes which enable monitoring the
possible changes of probability distributions over time. These
so-called monitoring processes are bivariate functions of time and
position at the measurement scale, and in particular be used to test
the null hypothesis of no change: one may then form Kolmogorov--Smirnov
or other type of tests as functionals of the processes. In Hjort and
Koning (2001) Cram??r-type deviation results were obtained under the
constancy null hypothesis for [bootstrapped versions of] such
``derived'' test statistics. Here the behaviour of derived test statistics is investigated under
alternatives in the vicinity of the constancy hypothesis. When
combined with Cram??r-type deviation results, the results in this
paper enable the computation of efficiencies of the corresponding
tests. The discussion of some examples of yield guidelines for the
choice of the test statistic, and hence for the underlying monitoring
process.Constancy of distributions: nonparametric monitoring of probability distributions over time
http://repub.eur.nl/pub/590/
Mon, 31 Dec 2001 00:00:01 GMT<div>N.L. Hjort</div><div>A.J. Koning</div>
In this paper we study stochastic processes which enable monitoring the
possible changes of probability distributions over time. These processes may
in particular be used to test the null hypothesis of no change. The
monitoring processes are bivariate functions, of time and position at the
measurement scale, and are approximated with zero mean Gaussian processes
under the constancy hypothesis. One may then form Kolmogorov--Smirnov or
other type of tests as functionals of the processes. To study null
distributions of the resulting tests, we employ KMT-type inequalities to
derive Cram\\'er-type deviation results for (bootstrapped versions of) such
tests statistics.