Love Canal is a 6.4 hectare tract, originally an uncompleted power canal, which was used for disposal of industrial chemicals in the 1940's and 1950's. The canal was later filled, and a school and residences were built on the site. Following a highly publicized discovery of contaminant migration from the canal in the 1970's, the school and nearby homes were demolished. Determination of the habitability of about 800 homes evacuated by Presidential order in 1980 in the surrounding Love Canal Emergency Declaration Area (EDA), was a highly visible environmental policy decision. The soil habitability study for the EDA was based on two‐sample comparisons of concentrations of eight trace organic indicators (Love Canal Indicator ChemicalsMdashLCICs) in EDA neighborhoods with selected comparison areas (CAs). The power of the statistical procedure (s) was an explicit criterion in the design of the study on which the habitability decision was based. Monte Carlo experiments were used to evaluate the relative performance of candidate statistical tests. The Monte Carlo simulations led to a recommendation for a pilot study to better identify the statistical characteristics of the soil chemical data. Analysis of the pilot study data, collected in the spring of 1986, suggested 1) that Monte Carlo experiments should be focused on univariate and multivariate Wilcoxon tests, 2) that a log normal mixture distribution provided reasonable fits to the observed data and could be used to evaluate a wide range of alternative hypotheses, and 3) that between‐laboratory variation would have to be explicitly addressed by the statistical test(s) and in experimental design. The test that was finally selected was a generalization of the Wilcoxon rank sum test for blocking by laboratory. The univariate (single chemical) and multivariate (multiple chemical) blocked Wilcoxon tests were applied to all combinations of EDA sampling areas and comparison areas using field data collected in the fall of 1987. For most comparisons and chemicals, the EDA sampling areas were found not to have significantly higher soil LCIC concentrations than the comparison areas. However, a few EDA sampling areas showed consistently higher concentrations for several LCICs. The multivariate test was particularly useful in avoiding the multiple comparison problem with respect to soil chemicals. Based in part on the results of the soil sampling study, the New York State Commissioner of Health, in September 1988, identified EDA neighborhoods that were suitable for rehabitation. Although the decision was made in a policy framework, the specific recommendations were consistent with the results of the statistical analyses.
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- Ecological Modeling