The decision rules commonly employed to determine the need for cleanup (including those cited in recent guidance) are evaluated below both to identify conditions under which they lead to erroneous conclusions and to quantify the rate that such errors occur. Their performance is also compared with that of other applicable decision rules.

We based the evaluation of decision rules on simulations. Results are presented as power curves. These curves demonstrate that the degree of statistical control achieved is independent of the form of the null hypothesis. The loss of statistical control that occurs when a decision rule is applied to a data set that does not satisfy the rule's validity criteria is also clearly demonstrated. Some of the rules evaluated do not offer the formal statistical control that is an inherent design feature of other rules. Nevertheless, results indicate that such “informal” decision rules may provide superior overall control of error rates, when their application is restricted to data exhibiting particular characteristics.

The results reported here are limited to decision rules applied to uncensored and lognormally distributed data. To optimize decision rules, it is necessary to evaluate their behavior when applied to data exhibiting a range of characteristics that bracket those common to field data. The performance of decision rules applied to data sets exhibiting a broader range of characteristics is reported in the second paper of this study.