Discussant's response to "Practical experiences with regression analysis";

              Discussant's Response to "Practical Experiences with Regression Analysis"
Peter R. Gillett
Formerly with Grant Thornton, U.K. Introduction
It has become a contemporary commonplace to characterise approaches to audit methodologies as either "quantitative" or "judgemental", and within the profession Price Waterhouse has traditionally been seen as occupying a place at the judgemental end of the spectrum. A study of the application of regression analysis within such a firm, examining its acceptability within a culture where quantified methods—statistical sampling, for example—have not been widely used, is of particular interest to those who have not yet found a place for it with-in their own armoury of audit tools.
My own firm, Grant Thornton, has adopted a structured audit approach incorporating a number of quantified audit methods, and yet even so regression analysis has not found favour. My discussion of this paper is no doubt coloured both by my own interest in quantified audit methods and by the limited accep-tance of regression analysis within the auditing profession in general. It may be helpful if I begin, then, by making explicit some of my prior expectations before reading this paper. They could be summed up, I suppose, by saying that I was looking to see what answers I might find to a number of difficulties that may be encountered by auditors seeking to apply regression analysis. These include:
1. Theoretical problems, such as:
• How are the calculations for regression analysis carried out?
• How is the validity of a regression model controlled?
• What do the various statistical terms associated with regression analysis mean?
2. Conceptual problems, such as:
• When is regression analysis applicable?
• What audit conclusions can be drawn from a regression analysis?
• How can regression analysis be integrated with other audit proce-dures?
• How are the results of regression analysis related to planning mate-riality?
• What degree of reliance can be placed on the results of regression analysis?
3. Practical problems, such as:
• Can plausible models be built within acceptable audit timeframes?
• Is sufficient client data available for regression analysis to be car-ried out?
170