The work of professionals in industrial hygiene and allied disciplines such as environmental health can substantially benefit from use of airborne exposure study design and data analysis methodologies that are based in mathematical statistics and probability theory. It has been said that “the science of statistics deals with making decisions based on observed data in the face of uncertainty”.
This chapter discusses some major areas of industrial hygiene practice where statistical methods perform an important role; the need for statistically sound study designs for both experimental and observational studies; discussions concerning statistical methods used for occupational epidemiological studies; concerning estimating possible threshold levels and low‐risk levels for occupational exposures; the area of application for statistics that is of primary interest and receives most attention in this chapter. This is the estimation of occupational exposures to airborne contaminants and calculation of error limits for such estimates. Nine possible objectives of occupational exposure estimation are discussed, which have their own special requirements for study design strategies.
Occupational exposure study designs and related data analysis methods have come to be broadly called sampling strategies. These sampling strategies are plans of action, based on statistical theory used to determine a logical, efficient framework for application of general scientific methodology and professional judgment.
Some basic statistical theory relevant to occupational exposure data are presented. Distributional models are given that identify the contributions of various sources of variation to the overall (net) random error in occupational exposure estimates. The National Institute for Occupational Safety and Health (NIOSH) nomenclature for exposure data is first given and a model is given for the contributions of the various components of variation to the net random error in occupational exposure measurements (due to the measurement procedure used), and the model is extended for total error to include random and systematic variations in true exposure levels (over times, locations, or workers doing similar work).
Information is included on the mathematical characteristics of basic distributional models. This is the starting point for deriving sampling distributions of industrial hygiene exposure data taken by various sampling strategies. General properties of the normal distribution model are given and of the lognormal distribution model (both two‐parameter and three‐parameter). The adequacy of normal and lognormal distribution models for certain general types of continuous variable data (specifically occupational exposure measurements) is discussed. These data models are then used to apply special‐interest applications of statistical theory to occupational exposure study designs and specialized data analyses.
Basic principles of statistically sound study design and data analysis that apply to all industrial hygiene surveys, evaluations, or studies are presented.
Particular study designs to be used for collecting data to estimate individual occupational exposures and exposure distributions are given. This section first discusses the cornerstone concept of worker target populations and discusses in detail another important concept, the determinant variables affecting occupational exposure levels experienced by a target population, discusses exposure measurement strategies selected to measure a short‐ or long‐period, time‐weighted average (TWA) exposure of an individual worker on a given day. Both practical and statistical considerations are discussed for long‐term and short‐term exposure estimates. Lastly, exposure monitoring strategies are presented for measuring multiple exposures (e.g., multiple workers on a single day, a single worker on multiple days, or multiple workers on multiple days). Eight possible elements of monitoring programs are discussed in detail, with examples of both exposure screening and exposure distribution monitoring programs.
The last section presents specialized applied methods for formal statistical analysis of occupational exposure data generated by the study designs discussed earlier.
