Estimating long-term exposures from short-term measurements.

Many health problems are related to chronic exposure of individuals to pollutants in the environment. The level of exposure of a specified population is typically represented by the mean level of exposure of the population, the variation in exposure between individuals within the population, and levels of exposure for selected percentiles of the population, such as the 50th and 98th percentiles. However, the day-to-day level of exposure for individuals varies, and direct measurement of total exposure for long periods of time is impractical. The problem is to estimate the quantities listed above using incomplete sampling of the time period of interest. This paper looks at the effect of using estimates of long-term exposure for individuals on estimating the exposure distribution of the population. A simple and apparently robust estimate for the upper percentiles of the exposure distribution is proposed. Problems related to estimating an individual's long-term exposure, including sample size, are also discussed. The length of time defined as long-term in this paper is one year; however, the results are generalizable to any period of time desired.