Comparisons of Low Concentration Measurement Capability Estimates in Trace Analysis: Method Detection Limit and Certified Reporting Limit.

摘要

Two large data sets were obtained over a four-day period for graphite furnace atomic absorption spectroscopic measurement of copper (Cu) and reversed-phase high performance liquid chromatographic determination of dinitrobenzene (DBN) at a number of concentrations near the lower limit of measurement. Low concentration measurement capability estimates for each analyte were obtained. For DNB, analytical variance was found to be homogeneous over the concentration range examined and MDL estimates were independent of concentration over the range of concentration examined. MDL estimates varied by as much as a factor of three from day to day, emphasizing the uncertainty in these estimates. CRL estimates varied to about the same extent and were numerically quite similar to MDLs when equivalent alpha and beta risks were used. For Cu, analytical variance was found to be proportional to concentration. Thus CRL estimates were very dependent on the concentration range examined. MDLs were less sensitive to this problem. Recommendations regarding the choice of target reporting limits for the CRL protocol were made. The influence of risk assumptions on both MDL and CRL estimates was examined and recommendations for modifications to both procedures made to incorporate an operational Beta-risk appropriate to the problem at hand. A case was made for using outlier tests to edit data used to estimate low concentration measurement capabilities. (AW)