Is measurement uncertainty from sampling related to analyte concentration?

Abstract

The contribution of sampling to the combined uncertainty of measurement is assessed using a combination of literature review and experimental determination of sampling variability in a range of foodstuffs in order to determine whether there is a consistent relationship between analyte level and proportion of variation attributable to sampling. Experimental determinations used the duplicate method, an economical method of assessing the relative contributions of sampling and analytical variability to the overall variance of results. The experimental work covered sampling of retail foodstuffs. 101 estimates of between-target, between-sampling, and within-sample variance were obtained. It is shown for the first time that sampling variance across the food sector appears to follow a Horwitz-like relationship sufficient to provide estimated between-sample standard deviation to within approximately an order of magnitude. The results from different methods of data processing for sampling uncertainty experiments are also compared. It is shown that for the data sets obtained experimentally in this study, log-transformation is of minor importance while the use of robust statistical methods can have greater but less predictable effects on estimated sampling variance.