Some Statistical Principles in Asbestos Measurement and Their Application to Dust Sampling and Analysis

摘要

Basic statistical principles of deterministic and stochastic variability are examined for their application to asbestos sampling and analysis methods, especially the approved and draft ASTM sampling and analysis methods for asbestos in dust. The sensitivities of these methods are shown to be inadequate for reliably determining the presence of asbestos structures in dust at 1000 s/cm~2 using currently-required counting protocols. Appropriate sensitivities can only be attained by counting impractically large numbers of grid openings. The current ASTM practice of stating detection limits for asbestos counts based on the confidence intervals for Poisson-distributed variables is shown to be wrong. Basic physical and statistical considerations show that detection limits are almost never appropriate for these methods, except in the presence of irremediable contamination of the sampling/analytical process.