Chemometrically guided sample composition for fast screening of trace metals in water samples

Abstract

A new strategy for building up composite samples, using experimental design matrices, useful in environmental and water quality control screening studies, is presented. While fixing the number of experiments to be carried out this chemometric strategic approach provides us not only with unbiased estimates of the true population mean, but also with analytical objects allowing the direct evaluation of the original individual sample specimens. This additional information has semiquantitative character but it is obtained without the need for analyses in the original samples nor retesting of positive composite samples. In this way, the main limitations of the classical concept of sample composition can be overcome while keeping all its advantages. In screening studies, frequently only a few of the original sample specimens are actually different from the background level (e.g. polluted hot spots). If it can be reasonably assumed that the number of actual anomalous or contaminated specimens, as compared to the total number of specimens, is low (effect sparsity or Pareto principle), then supersaturated experimental designs can be advantageously used to design the composite samples. This will provide the best conditions for this new strategy.