Practical experiences with regression analysis;

              7
Practical Experiences with Regression Analysis David A. Scott
Price Waterhouse, Canada
Wanda A. Wallace
College of William and Mary
Abstract
Price Waterhouse has conducted a field experiment on the application of regression analysis, involving the launching of new software, micro-based train-ing, and initial modeling for audit use. While the phases of the experiment included alpha and beta testing of the software as described herein, the core of the experiment involved field applications of regression analysis by engagement teams. Their experiences and reactions are described, as are the future plans of the firm. Experiences in prior field applications are likewise shared, to illustrate both the context in which regression analysis has been used and the nature of inferences drawn, as well as the statistical profile achieved in modeling. Insights are gained as to the tool's feasibility, time demands in its application, and per-ceptions of users.
Introduction
Over the years, a number of papers have appeared suggesting the benefits of using regression analysis as an analytical audit tool for risk identification and error detection. In some cases the authors have described individual applications of the technique. For example, Campbell and Rankin [1986] described the use of regression analysis to develop expectations of sales in a manufacturing com-pany, Kask [1979] covered an application to identify out-of-line energy costs for a group of hospitals, and Akresh and Wallace [1981] discussed a public util-ity application. Others [Knechel, 1986 and Wilson and Colbert, 1989] have reported that regression analysis, compared with alternate analytical procedures, is a more accurate tool for identifying errors of varying sizes and patterns seed-ed into simulated data.
Despite these purported conceptual advantages from using regression analy-sis, Deloitte & Touche is the only major accounting firm that seems to have used it regularly, in sampling applications [Stringer, 1975 and Stringer and Stewart, 1986—referred to as STAR]. While Price Waterhouse has had field applications using regression analysis since 1979, the scope of application has not been pervasive throughout the World Firm, for a number of reasons detailed later in this paper. Overall, as has been reported by Daroca and Holder [1985] and Spires and Yardley [1989], the use of regression analysis and other advanced quantitative procedures by audit teams, across firms, has been rela-
141