HYPOTHESIS-TESTING WITH VALUES BELOW DETECTION LIMIT IN ENVIRONMENTAL-STUDIES

Values below a specified detection limit are a common occurrence in en vironmental measurement which complicates statistical analysis. This p aper addresses the problem of how these data may be included in hypoth esis testing and demonstrates a solution that uses a regression model in the SAS statistical package with the ability to accommodate left-ce nsored data. To illustrate, the LIFER-EG procedure is applied to data for an analyte (tin) found in fish tissue during a multifactorial bioa ccumulation study involving two experimental sites, three experiments at each site, and sampling on multiple days. The results of this analy sis are presented, indicating evidence of bioaccumulation but no signi ficant differences between sites or experiments. While regression meth ods to accommodate left-censored data are not new, their use in enviro nmental studies does not appear to be widely practiced. The strengths and limitations of this approach are discussed.