Comment on 'Empirical versus modelling approaches to the estimation of measurement uncertainty caused by primary sampling' by J. A. Lyn, M. H. Ramsey, A. P. Damant and R. Wood.

Recently, Lyn et al. (Analyst, 2007, 132, 1231) compared two ways of estimating the standard uncertainty of sampling pistachio nuts for aflatoxins--a modelling method and an empirical method. Their case study used robust analysis of variance (RANOVA) to derive the uncertainty estimates, highlighting a substantial difference between the two: the estimate of sampling uncertainty derived from the modelling method was six-fold greater than that using the empirical approach (cf. 136% and 22.5%, respectively, when expressed as relative standard deviations (RSDs) at 68% confidence). A further analysis of this case study is reported here and suggests that the estimation uncertainty during RANOVA in the empirical approach could account for this difference.