Seminar Notice
Statistical Laboratory
Iowa State
University
DATE AND TIME: Wednesday, January 19, 2005, 4:10
p.m.
PLACE: 319 Snedecor
SPEAKER: Daniel
Nordman, Department of Statistics, University of Wisconsin – La
Crosse
TITLE: Developing Empirical Likelihood Under Long-Range
Dependence
ABSTRACT
This talk discusses the development of
empirical likelihood (EL) methods for linear time series which exhibit
long-range dependence or strong forms of dependence. For weakly dependent or
short-range dependent time processes, Kitamura~(1997) has proposed a version of
EL (blockwise EL) based on data blocking techniques that have successfully lead
to other nonparametric likelihoods, like the block bootstrap.
This
talk explains why blockwise EL, like the block bootstrap (Lahiri, 1993), does
not extend easily to long-range dependent time processes. One problem is that
blockwise EL ratios involve “`block adjustment” factors to ensure limiting
chi-square distributions and these adjustments are simple to formulate under
weak dependence but quite complicated under strong dependence.
As
an alternative to data blocking, we introduce a new version of empirical
likelihood based on the periodogram (a data transformation) and spectral
estimating equations. Under both weak dependence and strong forms of
dependence, the method results in likelihood ratios which can be used to build
nonparametric, asymptotically correct confidence regions for spectral parameters
like autocorrelations and Whittle
parameters.
COFFEE: 3:45 p.m., 104
Snedecor Hall