Building Empirical Models for Data from Factorial Designs with Time Series Responses: Toward Fraud Prevention and Detection
Quality Engineering, 1996/1997
Posted: 22 Jun 2004
Abstract
This paper examines how four factors (audit fee, penalty, required testing, and internal control strength) affected the percentage of times fraud was committed and detected in an experimental economics study. A factorial experiment resulted in time-series responses in each cell, with the factors' effects and interactions possibly a function of time. We introduce novel techniques for empirical model building and analysis for this type of data. The experiment indicated that the lowest fraud committed occurred when all four factors were high. The highest fraud detection occurred when internal control was low and the remaining factors were high.
Keywords: Experimental design, data exploration, contrast series, cumulative sum series, multiresponse
JEL Classification: M40
Suggested Citation: Suggested Citation