Uma comparação das distribuições inflacionadas para modelar dados duplamente limitados

Abstract

In the context of doubly bounded data, inflated distributions are widely used for cases where the data contain an excess of zeros and/or ones. There are several distributions in the literature to handle data in this domain, such as inflated beta (Ospina and Ferrari, 2010), inflated Kumaraswamy (Cribari-Neto and Santos, 2019), and inflated unitary gamma (Silva, 2023). In this work, we will present the most commonly used probability distributions in the context of modeling data distributed in the standard unit interval, but which may contain zeros and/or ones. Through a detailed study, via Monte Carlo simulation, we will identify scenarios according to the characteristics that are most suitable for each distribution.