Método empírico para cálculo do valor-e com auxílio do software R

Abstract

The present work was developed with the purpose of presenting an algorithm in R language which objective is, from a sample of a posterior distribution passed by the user of the code, to empirically calculate the e-value (Bayesian Evidence Value) based on a hypothesis also provided previously. In order to test the quality of the algorithm, it was validated in random samples generated from symmetrical and asymmetric distributions already known, in which it is possible to find out the exact value of the e-value through existing functions. With the help of the rjags package, the work done was also exemplified by testing hypotheses about parameters of a multiple Bayesian linear regression model and about the two parameters of a Normal distribution. Finally, it is emphasized that the code presented here has the potential to become a library of the R language at some later time.