Interlaboratory proficiency-testing programs: a computer model to assess their capability to correctly characterize intralaboratory performance.

Abstract

Abstract We developed a computer model of an interlaboratory survey program to study the ability of proficiency testing (PT) programs to detect intralaboratory errors (total, random, and systematic). It uses a base interlaboratory PT population of 400 laboratories and one test laboratory each with uniquely defined intralaboratory characteristics, i.e., mean, standard deviation (SD), and bias. A gaussian random-number generator uses these parameters to simulate 401 test results analogous to the analysis of one PT sample by each laboratory. The test laboratory's intralaboratory error is expressed as various combinations of bias and coefficient of variation (CV); its simulated survey result is evaluated by a performance criterion derived from the group statistics. To eliminate statistical artifacts, the computer model repeats the complete simulation process 400 times and determines the percentage of the test laboratory's results that fail to meet a specified performance criterion. The computer model can use assigned values or actual intralaboratory data.