Statistical Basis for Laboratory Performance Evaluation Limits

摘要

The U.S. Environmental Protection Agency (EPA) conducts studies to evaluate the performance of drinking water and wastewater laboratories that analyze samples for major EPA programs. The studies involve sample concentrates which the participating laboratories dilute to volume with reagent water. Analytical results are evaluated by comparison with confidence limits. From completed studies the characteristics of data representing proper laboratory operation are estimated using robust procedures on the data from EPA and state laboratories. From such historical statistics and their related true sample concentrations, weighted linear regressions are developed and used to predict the statistics for subsequent studies. If these regression predictions are similar to the statistics actually calculated from the current study data, the evaluation limits are calculated from the regression statistics for the sake of independence and to force a smooth relationship between the limits at different sample concentration.