Premium
ANALYZING CENSORED WATER QUALITY DATA USING A NON‐PARAMETRIC APPROACH 1
Author(s) -
She Nian
Publication year - 1997
Publication title -
jawra journal of the american water resources association
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.957
H-Index - 105
eISSN - 1752-1688
pISSN - 1093-474X
DOI - 10.1111/j.1752-1688.1997.tb03536.x
Subject(s) - estimator , statistics , log normal distribution , kaplan–meier estimator , mathematics , monte carlo method , survival analysis , parametric statistics , transformation (genetics) , delta method , cumulative distribution function , econometrics , computer science , probability density function , biochemistry , chemistry , gene
In the analysis of water quality data, samples with concentrations reported below the limit of detection (LOD) are referred to as Type I censored on the left. A variety of procedures have been proposed for estimating descriptive statistics from left‐censored data. Usually, the estimation is carried out by either replacing the LOD with a constant between 0 and the LOD, or assuming the data follow a normal or lognormal distribution. In this paper, a simple transformation is proposed to convert multiple left‐censored water quality data to right‐censored data. The transformed cumulative distribution is similar to a survival function, and enables use of survival analysis techniques for left‐censored data. In particular, the product limit method (Kaplan‐Meier estimator) is applied to estimate descriptive statistics from the transformed data. The performance of the Kaplan‐Meier estimator is compared with maximum likelihood, probability plotting, and substitution methods by Monte Carlo simulations. The Kaplan‐Meier estimator performs as well as or better than these more familiar methods. Finally, the Kaplan‐Meier estimator is used to analyze some priority pollutant data collected in sediment from the central basin of Puget Sound.