STATISTICAL ANALYSIS OF NONDETECTS IN BIOGEOCHEMICAL DATA

Measurements of trace chemicals in water, rocks, soils, and living organisms frequently result in values reported only as less than the laboratory detection limit (“less-thans” or “nondetects”). In biogeochemical studies, a common practice for interpreting these data is to substitute one-half the detection limit and perform traditional statistical tests. This overly-simplistic practice results in significant errors in interpretation. Groups may be declared different that are not, for example. Methods for interpreting right-censored (“greater-than”) data in medical and industrial statistics are available for estimating summary statistics, hypothesis testing, and regression, but have rarely been applied to biogeochemical data. Once adapted for left-censored “less-than” data, these methods return results that are unequivocal, powerful, and accurate -- unlike results based on fabrication of values not originally in the data set. Illustration of the use of these methods for biogeochemical data are summarized from the author’s newly-released textbook "Nondetects And Data Analysis: Statistics for Censored Environmental Data", published by John Wiley