Title: A comparison of methods to deal with observations below a limit of detection
Authors: Maria Polidoro - ESTG|P. Porto and CEAUL|FCUL (Portugal) [presenting]
Abstract: In many applied settings, often due to equipment limitations, measures below a certain limit of detection (LOD) cannot be obtained. Earlier approaches addressed this issue by omitting the observations below the LOD or by imputing them using values such as the LOD/2. However, these strategies can lead to biased estimates of the quantities of interest. Over the last decade or so, several approaches, both within the classical and Bayesian paradigms, have been proposed to properly handle observations that fall below the LOD. The goal is to compare several of these proposed methods via an extensive simulation study.