Regression Analysis as a Means of Determining Audit Sample Size: A Reply.

Abstract The original article by the authors was designed to explore the use of regression techniques in an audit environment and to provoke interest in and discussion of alternative approaches to the use of these models. There are, several major issues that must be resolved before the applicability of regression models to auditing can be ascertained fully. Professors Kinney and Bailey have articulated further some of these concerns. Their expansion of the authors' discussion indicates that part of the authors' purpose has been attained. An auditor would most likely use several variables in constructing the regression equation. The regression results were tested for the common problems associated with applications of regression analysis. The Durbin-Watson test was used to test the significance of serial correlation in the residuals. The correlation matrix for the four regressors did indicate multi-colinearity, with correlations in some cases as large as .39. However, omission of any one of the four predictors resulted in substantially larger prediction errors. The major overlap in predictors appeared between the economic series and the growth series. In an audit situation, the appearance of an incorrect account requires the auditor to conduct further investigation to determine whether the reported balance or the auditors estimate of such balance is correct.