2017-03-24

Li Chen, PhD, Department of Biostatistics, University of Kentucky

Estimation and Comparison of Environmental Exposure Distributions in Presence of Detection Limits

Epidemiologic studies investigating the relationship between exposure to environmental chemicals and cancer have been hampered by biases in measuring environmental chemical concentrations that fall below detection limits (DLs). The data subject to DLs present particular challenges for the estimation and comparison of environmental exposure distributions. Although several parametric and nonparametric methods have been previously proposed to address these two problems, there are critical limitations. First, current methods can lead to markedly biased results when the exposure level and DL are associated. Second, although nonparametric methods have drawn increasing attention in recent years due to their robustness, they are simple tweaks of methods for right-censored survival data and thus are not epidemiologically meaningful and may be inefficient for data subject to DLs. In this talk, we propose kernel-smoothed nonparametric estimation and testing methods for the data subject to DLs without imposing any independence assumption between the exposure level and DL. Simulation studies demonstrate that the proposed methods are valid regardless of whether the exposure level and DL are associated or not and can be more efficient than current methods. A colon cancer study is provided for illustration.

Stay connected TwitterFacebookLinkedInYouTubeInstagram