| ||Format||Pages||Price|| |
|PDF (356K)||16||$25||  ADD TO CART|
|Complete Source PDF (3.1M)||162||$55||  ADD TO CART|
This is the second paper of a study in which decision rules commonly employed to determine the need for cleanup are evaluated both to identify the conditions under which they may lead to erroneous conclusions and to quantify the rate at which such errors may occur. In this paper, performance is evaluated for such rules when applied to data exhibiting a range of characteristics commonly exhibited by environmental data. Results are reported for simulations involving data exhibiting normal distributions, lognormal distributions, and P-lognormal distributions (lognormal distributions with additional, finite density at zero). Some of the data sets employed were also censored.
Results indicate that none of the decision rules commonly applied to environmental data provide the advertised statistical control over the complete range of characteristics commonly exhibited by field data from hazardous waste sites. However, rules derived for normally distributed data or (modified from those derived for such data) appear to provide reasonable control of error rates that, appropriately, improve with increasing sample size. Also, because the advertised statistical control is seldom realized, it does not appear valid to prefer tests for which such formal control is inherent. Decision errors may best be controlled by matching selected decision rules to the observed characteristics of data.
risk assessment, site assessment, statistics, decision error, error rates, hypothesis tests, confidence limits
President, Aeolus Environmental Services, Albany, CA
Project Manager, ICF Kaiser Engineers, Research Triangle Park, NC
Van Landingham, CB
Project Manager, ICF Kaiser Engineers, Ruston, LA