Abstract Presents a reply to Bart Ward's commentary on the assessment of prior distributions for applying Bayesian statistics in auditing. Assumption regarding the problem of obtaining prior distributions of good quality; Fallacy of Ward's reasoning.
Abstract The article discusses assessing prior distributions for applying Bayesian statistics in auditing. The results of this study suggest that auditors are willing to specify information from which prior distributions can be constructed. The prior distributions which were obtained, had most of the probability concentrated on small amounts of error, but there was considerable variability among them. It was found that there were substantial inconsistencies in the way some auditors specified information about the prior distributions. These inconsistencies and the variability among the prior distributions indicate that at least some of these distributions do not accurately reflect the auditors' beliefs about audit populations. Therefore, it is concluded that auditors should proceed with caution in relying on their prior distributions. To apply Bayesian techniques, the auditor subjectively evaluates the non-sampling evidence and expresses his belief about the audit population as a prior probability distribution. A likelihood function is then obtained by statistically evaluating the sample result.