Proficiency Test Data, Population Characteristics of

Data from the measurement of silicate, acenaphthylene, and DDT in the marine environment are used to examine the Cofino model. This is a model for the inference of population characteristics from experimental data using uncertainties. It uses mathematical principles taken from quantum mechanics and provides an alternative to robust statistics for obtaining the population characteristics of proficiency test (PT) data. The Cofino model identifies clusters of values within a dataset that exhibit a high level of agreement and for each cluster calculates the mean, variance, and the percentage of the whole dataset. The model uses all data, including left-censored values, and can cope with bi- or multimodal distributions, skewed, or tailing data. The model is based on a normal distribution when applied to PT data, but the principles it uses can be applied to other distributions and data sources. Keywords: population characteristics; quantum chemistry; population measurement function; left-censored values; Cofino model; interlaboratory studies; proficiency testing