Comparison of solid sampling spectrochemical methods by means of multivariate statistics and information theory

Abstract

Interlaboratory studies and method comparison studies can usually be evaluated by using univariate statistical procedures. Alternatively, methods of multivariate data analysis can be applied to evaluate the reliability of methods more generally and simultaneously for all the elements determined by multicomponent analyses. The investigations described here were carried out to demonstrate the efficiency of multivariate and information theory evaluation techniques by comparing several multi-element solid sampling methods. Cluster analysis and PCA (principal component analysis) were used to compare the results of ETV-ICP-OES, solid sampling AAS, TXRF, dc arc spectrometry and dc arc spectrography by multivariate methods. The results correspond significantly to the information theory evaluation according to Kerridge–Bongard’s inaccuracy model of information, where both the precision and trueness of the analytical results are taken into account.