The variability introduced by various subsampling techniques

Publisher Website
Google Scholar
Abstract
Summary: Subsamples of a granular material contain a finite number of grains. Practical methods of subsampling do not select these grains on an entirely random basis. Both of these features contribute to the observed variability between subsamples. As an example, the theoretical coefficients of variation attributable to statistical variation between 10‐g random subsamples taken from four different (<2 mm) air‐dry soils have been calculated. These are calculated to give CVs of between about 0.002 and 1 per cent between particle size analysis results. In an actual comparison between four different subsampling methods, CVs of between 4 and 18 per cent were observed in the proportion of coarse sand (2–0.2 mm) in 15‐g soil subsamples. Rotary sub‐sampling and chute splitting were found to be more reliable methods of subsampling than cone and quartering or random spoon sampling. Variability between subsamples is shown to depend not only on subsampler design, but also on the precise details of the procedure adopted, and the physical properties of the sample itself.