A. Ochigbo

Publication Date

2014

Document Type

Dissertation/Thesis

First Advisor

Polansky, Alan M.

Degree Name

M.S. (Master of Science)

Legacy Department

Department of Statistics

LCSH

Auditing--Statistical methods; Sampling (Statistics); Statistics; Accounting

Abstract

A state government agency wants to determine if a sample verification procedure, used in auditing accounts to detect fraudulent charges, reduces the error of estimating the unknown mean amount of fraud per transaction. The procedure consists of checking whether the covariates of an audit variable, whose mean is known, is within a 100(1-alpha) % confidence interval computed on the observed sample. In this study we use computer based simulations to explore the effect that the procedure has on the error in estimating the mean. We concentrate on the bivariate normal distribution and on a normal bivariate normal mixture. Numerical results are presented that compare the estimated error for estimating the unknown population mean using the implementation of the sample acceptance algorithm and the standard method based on simple random sampling. The proposed method reduces the estimation error for multivariate normal distribution but can have the opposite effect for the non-normal distribution. Another approach based on the well known method of control variates produces similar results without the need to reject potential samples.

Comments

Advisors: Alan M. Polansky.||Committee members: Lei L. Hua; Peng Shi.

Extent

113 pages

Language

eng

Publisher

Northern Illinois University

Rights Statement

In Copyright

Rights Statement 2

NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.

Media Type

Text

Download

Share

COinS