Otimização do tamanho amostral na análise da qualidade de sementes de soja: abordagem bayesiana

Abstract

To ensure the quality of the seed passed on to the producer, seed resellers have adopted an internal quality control. As recommended by the Rules for Seed Analysis (RAS), for lots above 60 bags, it is advisable to take samples from 30 bags, which are punctured using a nozzle, where this can cause dissatisfaction or rejection by the customer. In addition, the greater the amount sampled, the greater the cost and waste generated from the analyzes. Therefore, studies are needed to minimize the number of punctured bags with the withdrawal of samples, without jeopardizing decisions regarding the usefulness of the analyzed batch. In order to make decisions based on the sample, the inference process is used, the Bayesian theory allows, by treating the parameter of interest at random, a more realistic interpretation of the studied phenomenon. Given these facts, the present study aimed to verify, using the Bayesian approach, with which sample size one can infer about the germination percentage of soybean seeds, without changing the decision criterion as to whether or not to accept the analyzed lot. For the experiment, the three main soybean seed suppliers in 2018 were selected, from a reseller located in the city of Alfenas. In the analysis, non-informative textit priori and two data sets were used as informative textit priori. To assess the effect of reducing the sample, of the 30 bags analyzed, 5000 subsamples were selected at random for each sample size (ns = 28, 26, 24, 22, 20, 18, 16, 14, 12, 10, 8, 6, 4). The decision to reject the lot or not was based on the limits and the amplitude of the 95 % credibility interval and the Bayes Factor log. In view of the results, it was observed that the use of the informative textit priori, presented a greater reduction in the sample size in comparison with the use of the non-informative textit priori, for most lots. It can be concluded that using a sample size greater than or equal to 14 bags, the decision made compared to the use of a sample of size 30 bags does not change, tends to reduce the dissatisfaction on the part of the producer, as well as the reduction of expenses and costs. waste generated from the analyzes.