Text

theses

Exploratory data analysis (EDA), which originated centuries ago, is a data analysis approach that emphasizes pattern recognition and hypothesis generation from raw data. It is suggested as the first step of any data analysis task for exploring and understanding data, and has been applied in many disciplines such as Geography, Marketing, and Operations Management. However, even though EDA techniques, such as data visualization and data mining, have been used in some procedures in auditing, EDA has not been employed in auditing in a systematical way. This dissertation consists of three essays to investigate the application of EDA in audit research. The study contributes to the auditing literature by identifying the importance of EDA in auditing, proposing a framework to describe how auditors could apply EDA to auditing, and using two cases to demonstrate the benefits that auditors can gain from EDA by following the proposed framework. The first essay identifies the value of EDA in auditing and proposes a conceptual framework to identify EDA’s potential application areas of EDA in various audit stages in both the internal and external audit cycles, describe how auditors can apply various EDA techniques to fulfill different audit purposes, and introduce a recommended process for auditors to implement EDA. In addition, this essay also discusses how EDA can be integrated into a continuous auditing system. The second essay examines the use of EDA in an operational audit. Traditional EDA techniques, such as descriptive statistics, data transformation, and data visualization, are applied in this credit card retention case. Descriptive statistics can reveal the distribution of the data, and data visualization techniques can display the distribution in an effective way so that auditors can easily identify patterns hidden in the data. By integrating these EDA techniques in audit tests, many critical risky issues, such as negative discount, inactive representatives, and short calls, are detected. With the rapid development of modern data analysis, EDA techniques have been greatly enriched. Besides the traditional EDA techniques, some data mining techniques can also be used to fulfill EDA tasks. The third essay investigates the application of two data mining techniques on a Medicare dataset to assess fraud risk. Specifically, this essay utilizes clustering techniques to detect abnormal Medicare claims in terms of claim payment amount and Medicare beneficiaries’ travel distances and hospital stay periods, and applies association analysis to analyze doctors’ diagnoses and performed procedures to identify abnormal combinations. In summary, this dissertation attempts to contribute to auditing literature by identifying the value that EDA can add to auditing, and illustrating some applications to demonstrate how auditors can benefit from EDA.

ETD_5890

Ph.D.

Includes bibliographical references

Includes vita

by Qi Liu

doi:10.7282/T3CC129J

ETD doctoral

The author owns the copyright to this work.