CONFLICTING PERSPECTWES ABOUT DETECTION LIMITS AND ABOUT THE CENSORING OF ENVIRONMENTAL DATA¹

Abstract

Current conventions for reporting analytical results from environmental samples brings the objectives of laboratory scientists into conflict with those of environmental scientists. The objective of chemical analyses is to provide estimates of the true composition of samples. Reported results must reflect the analytical uncertainty. Current conventions require left‐censoring of those results below the Limit of Detection. The objective of statistical interpretation of environmental data is to provide estimates of the characteristics of ecosystems. Such statistical analyses are often confounded by left‐censoring of analytical results. We review the different points of view and propose a compromise which recognizes these conflicting perspectives.

Optimization of automated gas sample collection and isotope ratio mass spectrometric analysis of δ13C of CO2 in air

“…Thus, our objectives were (1) to apply digital communication protocols between the sampling unit and the control computer to store status information from the sampling unit in order to eliminate the potential error of sample misidentification, cf. 20 (2) to increase the precision and reliability of the IRMS measurements for CO 2 in air samples and optimize sample preparation steps, and (3) to reduce the time required per IRMS analysis of a CO 2 in air sample in order to increase throughput in the laboratory and reduce storage times of the samples.…”

Determining the quality of taxonomic data

“…Error that exists in a consistently applied method is the systematic error, or bias, associated with that method or result (Smith et al 1988, Clark andWhitfield 1994). Bias is defined as statistical or method error caused by systematically favoring some outcomes over others and can be characterized as the degree of departure from a true value.…”

“…Accuracy is defined as the nearness of a measurement to some analytical truth (Fig. 2), or true value (Taylor 1988, Clark and Whitfield 1994, Taylor and Kuyatt 1994, Random House 1996. The key to determining the accuracy of any measurement or measurement system is specification of the true value, which often can be the most difficult aspect of the process.…”

New reporting procedures based on long-term method detection levels and some considerations for interpretations of water-quality data provided by the U.S. Geological Survey National Water Quality Laboratory

“…A simple approach to statistical analysis of censored data is substitution with 0, the censoring level, or one-half the censoring level. Other more robust methods exist for handling data analysis of censored and multiple censored data (Helsel and Hirsch, 1992;Clark and Whitfield, 1994;Clarke, 1998). These methods can be used for data reported with the new reporting procedure and include distributional methods, such as maximum likelihood estimation and probability plotting procedures, and robust methods to extrapolate censored values that are based on data above the censoring limit.…”