Estimation of the Linear Model with Right-Censored Covariates

• Activity Number: 182
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Folefac Atem*+ and Rebecca A. Betensky
• Companies: Harvard University and Harvard School of Public Health
• Keywords: Right Censored ; Covariates ; MLE ; Model Misspecification ; Complete Case

Abstract

Researchers are often faced with the problem of randomly censored covariates .The simplest and most straightforward approach for dealing with such data is to remove variables with censored observations or delete all censored observations. The former leads to model misspecification while the latter leads to overestimation of standard error due to a loss in power. Substitution methods, such as replacing censored values with a function of the censored values, have been widely utilized in the literature. Little and Rubin proposed a complete case analysis; Richardson and Ciampi proposed imputing right-censored values with ; Schisterman and colleagues proposed imputing censored values by the sample mean of the complete case; Rigobon and Stoker proposed the maximum likelihood approach. Ibrahim and colleagues extended the maximum likelihood approach to handle cases with more than one variable subject to limit of detection. For this study, we used simulations to compare the performance of these approaches.