Confidence intervals for quantiles when applying replicated latin hypercube sampling and sectioning

Authors

Marvin K. Nakayama, Department of Computer Science

Document Type

Conference Proceeding

Publication Date

12-1-2012

Abstract

Latin hypercube sampling (LHS), which can be viewed as an extension of stratified sampling to multiple dimensions, is a variance-reduction technique that increases statistical efficiency by inducing correlation among the outputs. We use a method known as sectioning to develop confidence intervals for quantiles when applying replicated Latin hypercube sampling (rLHS). Both two-sided and one-sided confidence intervals are given, and the intervals are asymptotically valid. One application of the technique is for uncertainty and safety analyses of nuclear power plants.

Identifier

84879438174 (Scopus)

ISBN

[9781622761012]

Publication Title

Simulation Series

ISSN

07359276

First Page

12

Last Page

19

Issue

16

Volume

44

Recommended Citation

Nakayama, Marvin K., "Confidence intervals for quantiles when applying replicated latin hypercube sampling and sectioning" (2012). Faculty Publications. 17914.
https://digitalcommons.njit.edu/fac_pubs/17914