Editorial: Number of Samples Can Make it or Break it

Most scientific journals have check-lists used by editors to get the first cut of manuscripts. High on that list for most journals, and top on the list for this Journal, is "Number of Samples." If the number is not adequate, the paper is rejected without further consideration. If the number is not specified, the paper will be returned to the author with a "Revise before review" letter, indicating that the number of samples must be specified and must be adequate. Comprehensive numeric information in the Materials and Methods section of every paper should include (1) number of units collected, (2) number of units composited, if applicable, (3) number of samples prepared for analysis and analysed (this is n), and (4) number of analytical replicates per sample. Tables containing statistical information should specify n, either as a footnote to the table or in the table headings. Replicate analysis is required for each sample to assure the precision of the analytical procedure, but whether a sample is put through the instrument twice or 100 times, it is still one sample (n=1). At times, replicate analyses can constitute samples in and of themselves (i.e., n=number of replicates). However, this is usually only in methods studies or qa/qc studies, where the purpose was not to provide food composition data, but to validate a procedure or develop a standard. Under most circumstances, a paper with n=1, or 2, will be rejected. The single composite sample presents a dilemma for editors and referees. Founding editor Kent Stewart, and guest editor Will Rand, often lamented the lost opportunity from total compositing. Some manuscripts present outstanding sample plans with representative unit collection, but with a single sample prepared by pooling all the units. Regardless of the number of units contained in the composite, this too is only one sample, albeit a better one than a single analysis of a single unit. There is a bare minimum of samples, always some number greater than two, the lowest of which depends on valid statistical inferences. However, we have accepted papers where n=1, where sampling was opportunistic, e.g., endangered animal species killed by poachers, other animals or vehicles, and recovered by wildlife or conservation agencies. We have also rejected papers where n=10 because the number was not sufficient to address the hypothesis (e.g., cultivar or geographical differences). This issue of the Journal of Food Composition and Analysis provides good examples of the variety of acceptable sampling plans and the requisite numeric data for samples. Of course, the number of samples is not the only feature that determines the quality of the research, but is the first feature examined in the process of turning a submitted manuscript into published paper.