Tratamento de dados censurados em estudos ambientais

Abstract

Due to the inherent limitations of the analytical methods of measurement, environmental exposure data often present observations described as below a certain detection limit, also called left-censored data. Censored data directly interferes in almost all types of statistical analyzes, including descriptive parameters, hypothesis testing, confidence intervals, correlations and regressions. In this work, we investigated the performance of the main classes of methods from major publications available in the literature, considering their advantages and limitations. Some criteria for selecting the best method of dealing with censored data are presented.