GRÁFICOS DE CONTROLE PARA PROCESSOS DE CONTAGEM E UNITÁRIOS: UMA ABORDAGEM DE CEP CLÁSSICO E BAYESIANO

Abstract

This work presents a set of implementations of distributions that consider process monitoring for count or truncated at (0,1) variables, such as rates and proportions. Thus, motivated by the demand for control of non-normal (non-Gaussian) processes, often encountered in applications in the engineering and economics fields, for instance, we implemented some functions in the R software for greater ease of SPC in applied areas. For counting data, the Poisson, Poisson-Lindley, Poisson-Shanker and Poisson-Sujatha distributions were adopted, and for unit data, the beta, Kumaraswamy, unit-Lindley and unit-half-normal distributions, all of which are uniparametric and reparameterizable in terms of mean or median. With this work, we hope to contribute to the literature with the implementation of such distributions via classical and Bayesian inference (via Hamiltonian Monte Carlo), resulting in visualization of modified Shewhart control charts. As an illustration, we present an application to air relative humidity data in the city of Copiapó - Chile (most arid region on the planet), adopting the developed functions that are available in a repository on GitHub.