Abstract In a recent article, Edward B. Deakin and Michael H. Granof " demonstrated how regression analysis, coupled with Bayesian statistical procedures, can be used to provide the auditor with assistance in selecting those accounts for investigation that are most likely to result in significant audit findings. The article encourages audit model development and experimentation by independent auditors with the application of these and other techniques which make use of additional information. The independent auditor's ultimate concern with respect to a reported account balance of a client, is whether it is "fairly presented." Suppose that the auditor has evaluated the design of an internal control subsystem and has conducted tests of compliance of system operation with system design. Substantive tests include analytical review of significant ratios and trends and resulting investigation of unusual fluctuations and questionable items and tests of details of transactions and balances. Without the refinement, the sample sizes required after using regression analysis to revise priors often will be substantially smaller than the 163 required under classical, unrestricted random sampling.
Kinney et al. (Thu,) studied this question.