Page 141 - Statistics for Environmental Engineers

L1592_Frame_C15 Page 138 Tuesday, December 18, 2001 1:50 PM

the median value. If the median is above the MDL, draw a smooth curve through the plotted points and
estimate the median directly. If the median is below the MDL, extrapolation will often be justified on
the basis of experience with similar data sets. If the data are distributed normally, the median is also the
arithmetic mean. If the distribution is lognormal, the median is the geometric mean.
The precision of the estimated mean and variances becomes progressively worse as the fraction of
observations censored increases. Comparative studies (Gilliom and Helsel, 1986; Haas and Scheff, 1990;
Newman et al., 1989) on simulated data show that Cohen’s method works quite well for up to 20% censoring.
Of the methods studied, none was always superior, but Cohen’s was always one of the best. As the extent
of censoring reaches 20 to 50%, the estimates suffer increased bias and variability.
Historical records of environmental data often consist of information combined from several different
studies that may be censored at different detection limits. Older data may be censored at 1 mg/L while
the most recent are censored at 10 µg/L. Cohen (1963), Helsel and Cohen (1988), and NCASI (1995)
provide methods for estimating the mean and variance of progressively censored data sets.
The Cohen method is easy to use for data that have a normal or lognormal distribution. Many sets of
environmental samples are lognormal, at least approximately, and a log transformation can be used.
Failing to transform the data when they are skewed causes serious bias in the estimates of the mean.
The normal and lognormal distributions have been used often because we have faith in these familiar
models and it is not easy to verify any other true distribution for a small sample (n = 20 to 50), which
is the size of many data sets. Hahn and Shapiro (1967) showed this graphically and Shumway et al.
(1989) have shown it using simulated data sets. They have also shown that when we are unsure of the
correct distribution, making the log transformation is usually beneficial or, at worst, harmless.

References
Aitchison, J. (1955). “On the Distribution of a Positive Random Variable Having a Discrete Probability Mass
at the Origin,” J. Am. Stat. Assoc., 50, 901–908.
Aitchison, J. and J. A. Brown (1969). The Lognormal Distribution, Cambridge, England, Cambridge University
Press.
Berthouex, P. M. and L. C. Brown (1994). Statistics for Environmental Engineers, Boca Raton, FL, Lewis
Publishers.
Blom, G. (1958). Statistical Estimates and Transformed Beta Variables, New York, John Wiley.
Cohen, A. C., Jr. (1959). “Simplified Estimators for the Normal Distribution when Samples are Singly Censored
or Truncated,” Technometrics, 1, 217–237.
Cohen, A. C., Jr. (1961). “Tables for Maximum Likelihood Estimates: Singly Truncated and Singly Censored
Samples,” Technometrics, 3, 535–551.
Cohen, A. C. (1979). “Progressively Censored Sampling in the Three Parameter Log-Normal Distribution,”
Technometrics, 18, 99–103.
Cohen, A. C., Jr. (1963). “Progressively Censored Samples in Life Testing,” Technometrics, 5(3), 327–339.
Gibbons, R. D. (1994). Statistical Methods for Groundwater Monitoring, New York, John Wiley.
Gilbert, R. O. (1987). Statistical Methods for Environmental Pollution Monitoring, New York, Van Nostrand
Reinhold.
Gilliom, R. J. and D. R. Helsel (1986). “Estimation of Distribution Parameters for Censored Trace Level Water
Quality Data. 1. Estimation Techniques,” Water Resources Res., 22, 135–146.
Hashimoto, L. K. and R. R. Trussell (1983). Proc. Annual Conf. of the American Water Works Association,
p. 1021.
Haas, C N. and P. A. Scheff (1990). “Estimation of Averages in Truncated Samples,” Environ. Sci. Tech., 24,
912–919.
Hahn, G. A. and W. Q. Meeker (1991). Statistical Intervals: A Guide for Practitioners, New York, John Wiley.
Hahn, G. A. and S. S. Shapiro (1967). Statistical Methods for Engineers, New York, John Wiley.
Helsel, D. R. and T. A. Cohen (1988). “Estimation of Descriptive Statistics for Multiply Censored Water
Quality Data,” Water Resources Res., 24(12), 1997–2004.
Helsel, D. R. and R. J. Gilliom (1986). “Estimation of Distribution Parameters for Censored Trace Level Water
Quality Data: 2. Verification and Applications,” Water Resources Res., 22, 146–55.
© 2002 By CRC Press LLC