A Statistical Approach for Left-Censored Data: Distributions of Atmospheric Polychlorinated Biphenyl Concentrations near the Great Lakes as a Case Study

摘要

Polychlorinated biphenyl (PCB) congener concentrations were measured in atmospheric samples collected once every 12 days at six sites on the shores of the North American Great Lakes. These data have been obtained as part of the Integrated Atmospheric Deposition Network (IADN), which began in 1991. This data set now consists of ～2900 samples, each of which has been measured for ～80 PCB congeners. Some of these congeners are present at levels sufficiently high to be quantitated in almost every sample, while others are detected in fewer than two-thirds of the samples. These latter congeners represent an example of left-censored environmental measurements. This paper offers a simple approach to dealing with uncensored and censored atmospheric PCB concentration data based on a careful examination of the distribution function of the data and using the curve fitting power of the Solver feature of Excel.