The replication myth 2: Quantifying empirical sampling plus analysis variability

In the previous column, it was stated that in order to furnish a realistic estimate of the complete Total Sampling Error (TSE) + Total Analytical Error
(TAE), a Replication Experiment (RE) must always start “from the top”, i.e. it is the primary sampling which is to be replicated. The primary sampling may be field sampling, sampling at the industrial plant or it
can be sampling of any lot designated as the primary lot (examples follow below). In this fashion, and only in this fashion, will all sampling errors be represented (primary sampling, sub-sampling, mass reduction
errors associated with sample preparation
etc.) in ten (or more) individual manifestations—to which is added the total analytical error, TAE.