Optimization of Bayesian Multiple Comparison Tests dbayes and pbayes in R Code

• Main Author: Ramos, Carlos Henrique Oliveira
• Publication Date: 2019
• Other Authors: Araújo, Fernanda Alves, Andrade, Paulo César de Resende
• Format: Article
• Language: eng
• Source: Revista Semina: Ciências Exatas e Tecnológicas (Online)
• Download full: https://ojs.uel.br/revistas/uel/index.php/semexatas/article/view/36405
• topic: Bayesian tests; MCP. Analysis of Variance; Completely Randomized Designs; Testes Bayesianos; PCM; Análise de Variância; Delineamento Completamente Aleatorizado.; Probabilidade e Estatística

The experimental statistic uses multiple comparison procedures (MCP) to verify if there is a difference between the treatments under analysis. However, the presence of unbalanced data and the cases of heterogeneity of variance negatively influence the performance of the most used tests. The dbayes and pbayes tests were previously implemented in the context of completely randomized designs by one of the authors. These tests are valid for cases where assumptions of variance analysis are met or not, with or without balancing. The objective of this article is to optimize the Bayes function, in R code, that allows the performance of these tests. To validate the optimization, it compared the optimized code with the previous code and used three real situations: one considering all the assumptions, the other two with unbalanced data and with different numbers of treatments. The optimized Bayes function allows the dbayes and pbayes tests to perform well under conditions of assumption and balancing. These tests can be used satisfactorily in situations of non-compliance with the assumptions. In cases of unbalanced data, with a small number of treatments, the dbayes test presents a result superior to the Tukey-Kramer test