Publication detail

Priestley-Chao Estimator of Conditional Density

KONEČNÁ, K.

Original Title

Priestley-Chao Estimator of Conditional Density

English Title

Priestley-Chao Estimator of Conditional Density

Type

conference paper

Language

en

Original Abstract

This contribution is focused on a non-parametric estimation of conditional density. Several types of kernel estimators of conditional density are known, the Nadaraya-Watson and the local linear estimators are the widest used ones. We focus on a new estimator - the Priestley-Chao estimator of conditional density. As conditional density can be regarded as a generalization of regression, the Priestley-Chao estimator, proposed initially for kernel regression, is extended for kernel estimation of conditional density. The conditional characteristics and the statistical properties of the suggested estimator are derived. The estimator depends on the smoothing parameters called bandwidths which influence the final quality of the estimate significantly. The cross-validation method is suggested for their estimation and the expression for the cross-validation function is derived. The theoretical approach is supplemented by a simulation study.

English abstract

This contribution is focused on a non-parametric estimation of conditional density. Several types of kernel estimators of conditional density are known, the Nadaraya-Watson and the local linear estimators are the widest used ones. We focus on a new estimator - the Priestley-Chao estimator of conditional density. As conditional density can be regarded as a generalization of regression, the Priestley-Chao estimator, proposed initially for kernel regression, is extended for kernel estimation of conditional density. The conditional characteristics and the statistical properties of the suggested estimator are derived. The estimator depends on the smoothing parameters called bandwidths which influence the final quality of the estimate significantly. The cross-validation method is suggested for their estimation and the expression for the cross-validation function is derived. The theoretical approach is supplemented by a simulation study.

Keywords

kernel smoothing; conditional density; Priestley-Chao estimator; statistical properties; bandwidth selection; cross-validation method

Released

01.12.2017

Publisher

University of Defence, Brno, 2017

Location

Brno

ISBN

978-80-7582-026-6

Book

Mathematics, Information Technologies and Applied Sciences 2017, post-conference proceedings of extended versions of selected papers

Pages from

151

Pages to

163

Pages count

13

URL

BibTex


@inproceedings{BUT142655,
  author="Kateřina {Pokorová}",
  title="Priestley-Chao Estimator of Conditional Density",
  annote="This contribution is focused on a non-parametric estimation of conditional density. Several
types of kernel estimators of conditional density are known, the Nadaraya-Watson and the
local linear estimators are the widest used ones. We focus on a new estimator - the Priestley-Chao
estimator of conditional density. As conditional density can be regarded as a generalization of regression,
the Priestley-Chao estimator, proposed initially for kernel regression, is extended for kernel
estimation of conditional density. The conditional characteristics and the statistical properties
of the suggested estimator are derived. The estimator depends on the smoothing parameters called
bandwidths which influence the final quality of the estimate significantly. The cross-validation
method is suggested for their estimation and the expression for the cross-validation function is derived.
The theoretical approach is supplemented by a simulation study.",
  address="University of Defence, Brno, 2017",
  booktitle="Mathematics, Information Technologies and Applied Sciences 2017, post-conference proceedings of extended versions of selected papers",
  chapter="142655",
  howpublished="online",
  institution="University of Defence, Brno, 2017",
  year="2017",
  month="december",
  pages="151--163",
  publisher="University of Defence, Brno, 2017",
  type="conference paper"
}