Science and Judgment in Risk Assessment (1994)

•

Variability is (or can be modeled sufficiently precisely as though it were) is) either discrete or continuous. For example, albinos are many times more sensitive to sunlight than other members of the population, so a (dichotomous) discrete assumption might well be appropriate here. In contrast, because body weights vary continuously, the cancer risk per unit dose of a substance cannot be modeled dichotomously without the loss of much of information.

•

Variability is identifiable or unidentifiable. Albinism is a good example of identifiable variability, whereas the extent of a person's ability to detoxify a particular active metabolic intermediate might not be discernible without invasive testing, and hence is unidentifiable for most of the population.

•

Identifiable variability is dependent on or independent of additional variable characteristics that society deems salient. For example, some factors that cause genetic predisposition to the carcinogenic effect of chemicals are correlated with race, sex, or age. If society deems that those who are predisposed already deserve special attention because of the other factors, the importance of the variability is heightened. But some kinds of identifiable variability, such as body weight and phenylketonuria, are more "value-neutral" or are uncorrelated with any relevant characteristic.