You are here:
PROUCL VERSION 3.0
Citation:
Singh, A., R. Maichie, AND A. K. Singh. PROUCL VERSION 3.0. U.S. Environmental Protection Agency, Washington, DC, 2004.
Impact/Purpose:
The overall objectives of this task are to: 1) provide ORD state-of-the-science technical support and assistance to Regional staff; 2) facilitate the evaluation and application of site characterization technologies at Superfund and RCRA sites; and 3) improve communication among Regions and ORD laboratories.
Description:
The computation ofa (l-a) 100% upper confidence limit (UCL) of the population mean depends upon the data distribution. Typically, environmental data are positively skewed, and a default lognormal distribution (EPA, 1992) is often used to model such data distributions. The H-statistic based Land's (Land 1971, 1975) H-UCL of the mean is used in these applications.
Hardin and Gilbert (1993), Singh, Singh, and Engelhardt (1997,1999), Schultz and Griffin,1999, Singh et al. (2002a), and Singh, Singh, and Iaci (2002b) pointed out several problems associated with the use of the lognormal distribution and the H-UCL of the population AM. In practice, for lognormal data sets with high standard deviation (sd), a, of the naturallog-transformed data (e.g., 0" exceeding 2.0), the H-UCL can become unacceptably large, exceeding the 95% and 99% data quantiles, and even the maximum observed concentration, by orders of magnitude (Singh, Singh, and Engelhardt, 1997). This is especially true for skewed data sets of smaller sizes (e.g., n < 50).
The H-UCL is also very sensitive to a few low or high values. For example, the addition of a sample with below detection limit measurement can cause the H-UCL to increase by a large amount (Singh, Singh, and Iaci, 2002b). Realizing that use of the H-statistic can result in unreasonably large UCL, it is recommended (EPA, 1992) to use the maximum observed value as an estimate of the UCL (EPC term) in cases where the H-UCL exceeds the maximum observed value. Recently, Singh, Singh and Iaci (2002b), and Singh and Singh (2003) studied the computation of the UCLs based upon a gamma distribution and several non-parametric bootstrap methods. Those methods have also been incorporated in ProUCL Version 3.0. ProUCL Version 3.0 contains fifteen UCL computation methods; five are parametric and ten are non-parametric. The non-parametric methods do not depend upon any of the data distributions.