Uncertainty budget The following includes bias correction by dividing future concentration estimates by (1 + biâs) as in Technical Note 2.
Source
Component
Category
Inter-sampler
3.9% [= 4.4% / (1+.133)]
Type A
Bias Correction Uncertainty
0.97% [= 3.9%/
Type A
Ref Concentration Uncertainty
0.5% [1%
] , but not /(1 +
Combined (Relative) Uncertainty
)]
Type B
= 4.0%
Expanded (Relative) Uncertainty
(k = 3).
Notes: Here k = 3.0. A more accurate determination based on the chi-square quantile at 12 degrees of freedom and prob = 0.05, gives k = 2.97, which is consistent with conventional use of 3 as a conservative value. Again, an expanded (relative) uncertainty U means that with greater than 95% of future biascorrected estimates , true concentrations C are bracketed by:
at 95% confidence in the above evaluation experiment. Generally, a quality control program is required to ensure that the method remains stable following evaluation. Note that many methods (e.g., those based on sorbent tubes) employ personal sampling pumps, in which case normally a 5% (Type B) component representing sampling pump uncertainty would be included in the uncertainty budget. Note also that the inter-sampler component includes both analytical and sampling subcomponents. Further refinement of the inter-sampler component may perhaps be useful for improving a method, but is not needed for establishing confidence intervals around (true) measurand values. Note further that storage effects require estimating and inclusion in the budget if considered significant.
3/15/03
217
NIOSH Manual of Analytical Methods
