Page:NIOSH Manual of Analytical Methods - Chapter P.pdf/3

 type of aerosol sampler is characterized by specific particle collection characteristics, and some analytical methods (e.g. silica) may also exhibit particle size effects. Typically the issue of aerosol sampler bias is avoided or minimized in the industrial hygiene field by narrowing use to a specific aerosol sampler. For example, common industrial hygiene practice establishes a single sampler type, such as the 1.7 L/min 10-mm nylon Dorr-Oliver cyclone, for respirable dust sampling in a particular application. Sensitivity to other environmental factors, referred to in ISO GUM as influence variables, must be acknowledged. Suppose a sampler is sensitive to temperature changes that are impractical to measure in the field; i.e., sampler estimates are not temperature corrected. Then, suppose during method evaluation in the laboratory, measurement of this sensitivity is combined with knowledge of the expected temperature variation for a given field application. Putting together both would determine the uncertainty associated with the effect. Examples of the important effects of influence variables - such as wind velocity, temperature, pressure, and fluctuating workplace concentrations - on diffusive monitor uptake rates are common.

2.

ISO GUM ISO GUM presents several concepts. One of these calls for the identification of sources (labeled j = 1, 2,...) of uncertainty uj (standard deviation estimate components) in a measurement method and for their classification into Type A or Type B uncertainties. Type A uncertainty is one that has been characterized by a statistically sound approach. In this case,

is given by

, an unbiased estimate (with υj degrees of freedom) of variance

.

On the contrary, Type B uncertainty generally requires professional judgment. See Table 1 for examples of possible uncertainty components. A common example of Type B uncertainty is the conservative assignment (Chapter E) of a 5% relative standard deviation component (without error, i.e., with infinite degrees of freedom) as the random sampling pump uncertainty. As described in ISO GUM, such an assignment would be a result of sampling pump random errors that had a uniform distribution and fell within of zero with a probability “for all practical purposes equal to one”. Therefore, if it is judged that sampling pump variations are within these bounds, then the assignment of 5% as the relative standard deviation component is conservative. Other similar ways of handling Type B uncertainties are found in ISO GUM.

3/15/03

210

NIOSH Manual of Analytical Methods