RESUMO
Incremental samples (50, 100, and 500 g) were systematically collected from large shipments of copra meal pellets, copra cake, and palm kernel cake to study the distribution of aflatoxin B1 and evaluate adherence of distribution to the model, CV(2)is (EQ) = A + B/Mis (where CVis = coefficient of variation of the true concentration of aflatoxin B1 within the incremental samples; Mis = mass of the incremental samples; and A and B are constants). Also evaluated was the distribution of aflatoxin B1 among 1 kg composite samples, produced both by random combination of existing incremental samples and by collection of 1 kg composite samples (composed of 10 x 100 g increments) from additional batches of copra meal pellets and cottonseed cake. The efficiency of selected sample preparation (grinding and subdivision) procedures was compared, culminating in the development and description of a variety of sampling plans. The coefficient of variation (CV) among incremental samples varied from 0 to 38%, and was independent of incremental sample size. No significant difference (F-test, 5% significance level) was found between the efficacy of 4 sample preparation methods when these methods were applied to the commodities described above. Various sampling plans were evaluated with estimated CVs from 4.0 to 12.5%, for the aflatoxin B1 content of the composite samples.