Statistical methodologies for exposure assessment.
Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health, R01-OH-003628, 2009 Aug; :1-8
The proposed research was on the development of statistical procedures appropriate for analyzing exposure data, based on statistical models directly relevant for exposure assessment. The following topics were investigated: the univariate and bivariate lognormal distributions that describe exposure data, methodology for comparing several test methods or samplers, procedures to analyze samples that included values below the detection limit, and random effects models for describing variability among workers. Difficulties and limitations of some of the currently used statistical techniques were highlighted, and efficient alternatives were developed. Approaches based on the novel concepts of generalized p-values and generalized confidence intervals were successfully investigated for solving the proposed problems. The problem of assessing occupational exposure using the mean of a lognormal distribution was addressed using the concepts of generalized p-values and generalized confidence intervals. The resulting methodologies are applicable to small samples and they are easy to implement. Power studies and sample size calculations have also been discussed. The procedures have also been extended for the purpose of comparing two independent exposure populations, as well as for comparing the means of a bivariate exposure population, following the bivariate lognormal distribution. A model based "multiple imputation approach" was developed for analyzing exposure data with non-detects. The approach involves replacing the non-detects with randomly generated observations under the appropriate model, and then analyzing the data using complete sample techniques after adjusting for the imputation. The method has been described for lognormally distributed exposure data, and has been illustrated for computing prediction limits and tolerance limits, for setting an upper bound for an exceedance probability, and for estimating the arithmetic mean. In many industrial hygiene applications, a model with random effects is used to account for the variability within an exposure group. Under this setting, statistical methods have been proposed for setting upper limits on (i) the probability that the mean exposure of an individual worker exceeds the occupational exposure limit (OEL) and (ii) the probability that the exposure of a worker exceeds the OEL. The proposed method for (1) was obtained using the generalized confidence interval approach, and the one for (ii) was based on an approximate method for constructing one-sided tolerance limits. The methods are conceptually as well as computationally simple. The comparison of an alternative sampling device to an OSHA standard, or the comparison of two sampling devices, or that of several sampling devices and methods, is a problem that is frequently encountered while developing cheaper or more efficient exposure assessment strategies. A test procedure has been developed for making such comparisons. Performance of the test has been numerically investigated, and satisfactory performance has been noted. Preliminary work has been completed to investigate the NIOSH accuracy criterion based on the symmetric-range accuracy, for the quantification of measurement accuracy of exposure data. The symmetric-range accuracy of a sampler is defined as the fractional range, symmetric about the true concentration, that includes a specified proportion of sampler measurements. An explicit expression has been derived for the symmetric-range accuracy when the sampler measurements follow a normal or a lognormal distribution. Confidence limits have been proposed for the symmetric-range accuracy using the generalized confidence interval idea. An accurate and convenient approximation has also been developed for computing the confidence limit. The newly developed statistical methodologies have all been illustrated by applying them to the analysis of real exposure data. Computational algorithms and software codes have also been provided.
Statistical-analysis; Exposure-levels; Samplers; Analytical-processes; Models; Workers; Industrial-hygiene
Thomas Mathew, Department of Mathematics and Statistics, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250
Final Grant Report
NTIS Accession No.
National Institute for Occupational Safety and Health
University of Maryland, Baltimore