Imputación de Valores Faltantes de Variables de Razón y/o Proporción Mediante Modelos de Regresión Beta Cero-uno Inflados

Abstract

In the data collected in recent decades, it is common to encounter the presence of missing values for multiple reasons, which can lead researchers to use incorrect strategies to handle them, resulting in incorrect inferences. This problem is much more common in ratio and proportion variables, as most multiple imputation methods do not take into account the nature of these types of variables. In the literature, there are models for this type of variable, such as Beta regression models, which can be used to construct multiple imputation models. Given this need, the present work proposes to develop a function in the R software that uses multiple imputation methods with inflated zero-one Beta regression models to allow for the correct implementation of imputation for proportion and/or ratio variables.