Proper Sampling: A Precondition for Accurate Analyses

Abstract

There can be no discussion of sampling without an unambiguous problem definition and a clear formulation of the analytical task. The systematics of the series EN 45 000 stipulates that important sampling can only be carried out by surveillance laboratories which are authorized to draw conclusions from single samples to the whole.

In those cases where the analytical sample represents the entire population, sampling is inapplicable as a selection process. This is true for very small samples; the so-called integration error then becomes zero. In general, inhomogeneities resulting from the discrete structure of material, from the persistence and the periodicity of a process (autocorrelation), contribute to this integration error.

Strategies for the reduction of these error components are briefly discussed.