BIAS IN POPULATION ESTIMATES OF LONG-TERM EXPOSURE FROM SHORT-TERM MEASUREMENTS OF INDIVIDUAL EXPOSURE

A population's long-term exposure distribution for a specified compoun d is typically estimated from short-term measurements of a sample of i ndividuals from the population of interest. in this situation, estimat es of a population's long-term exposure parameters contain two sources of sampling error: the typical sampling error associated with taking a sample from the population and the sampling error from estimating in dividual long-term exposure. These components are not separable in the data collected, i.e., the value observed is due partly to the individ ual sampled and partly to the time at which the individual was sampled . Hence, the distribution of the data collected is not the same as the population exposure distribution. Monte Carlo simulations are used to compare the distribution of the observed data with the population exp osure distribution for a simple additive model. A simple adjustment to standard estimates of percentiles and quantiles is shown to be effect ive in reducing bias particularly for the upper percentiles and quanti les of the population distribution.