Calculation of decision levels and minimum detectable concentrations from method blank and sample uncertainty data — utopian statistics

Abstract

Calculation of meaningful minimum detectable concentrations (MDCs) and decision levels (DLs) requires proper evaluation of all uncertainties and biases of the measurement process as determined by appropriate blanks. Using a population of method blank results to estimate σblank is appropriate for calculation of MDCs since MDC is an a priori concept and not intended to be sample specific. Using a population of method blanks for calculation of sample specific decision levels (DLs) requires estimation of σblank from net counts or count-rate so that sample specific efficiency, recovery, decay correction, and other factors can be used. However, data has shown that DLs calculated as 1.65σsample are easier to implement and compare favorably with DLs calculated as the tσblank when a total propagated uncertainty (TPU) is reported by the laboratory and used to estimate σsample instead of just the counting uncertainty. DLs calculated in either of these ways resulted in no more than 5% false positives.