Abstract
Uncertainties in the collection and assessment of scientific information make ambient air quality standard setting difficult. Uncertainties occur in the estimation of the medical parameters under test due to the inherent random variability encountered in sampling the parameters. The most common method of dealing with random variability is statistical significance testing. The main caution offered in regard to that analysis is to avoid calling a nonsignificant result negative, unless the circumstances are such that the smallest effect which indicates likely harm to health could have been detected with sufficiently high probability. Uncertainties also play a crucial role in evaluating the implications that even statistically significant test results have for human health. A signal-detection model, developed to explain expert performance in evaluating the results of such diagnostic tests as X-rays, is presented as an analogy for the situation facing experts who are evaluating the implications of health data that is being considered for use in setting a standard. If criteria are too strict for accepting data as evidence of harm to health, then it is argued that, as a consequence, the decision process will not have sufficient ability to discriminate against false-negative results. False-negative results are those that incorrectly conclude there is no threat when, in fact, a particular level of pollutant is actually a threat to health.