Análise bayesiana do potencial de viabilidade de sementes de mamona por meio de análise radiográfica

Abstract

In this paper, we observed the potential feasibility of castor seeds from the visual analysis of its radiographs, made by three independent experts using Bayesian methods. We used 800 seeds of castor bean from cultivar IAC-226, harvested in 2009, identified individually, and divided into two samples of 400 seeds each. The first one was subjected to X-ray test without any preparation and the second sample pre-treated with the gaseous substance contrast. After that, using the relative frequency and Bayesian methods, we calculated the likelihood of the following possible outcomes seed which germinate or originated an irregular non standard seedling have been evaluated as potentially feasible. To find the Bayes estimators we used the Gibbs sampler, in which the samples of the conjoint distribution were generated using methods of computational simulation. The convergence of the chains was verified by tests of Geweke, Raftery and Lewis and Heidelberg and Welch. The simulation of different sizes of samples aimed to compare the estimates obtained using Bayesian methods to the values observed in the samples. The Bayesian method was efficient in the estimation of the parameters, which is close to those estimated by relative frequency. Comparing the samples of different sizes we did not observe significant differences among the results. We concluded that the Bayesian methods can be applied in a large spectrum of attempts to obtain good inferences in the science field of seed technology, since they can incorporate both quantitative and qualitative attributes such as the opinions of specialists. Furthermore, they can be used to study the viability of seeds of other species.