Research Product

As part of a larger program to examine the relationships between laboratory and ambient toxicity, data from a series of 28 marine microcosm experiments, which included both laboratory and field studies, were reanalyzed. The statistical methods applied to the data include log-linear regression models (general and robust) for richness and taxa count data (kingdom, phylum, class, order, family, genus, and species) and Gaussian linear models (general and robust) for community metrics diversity and eveness (Shannon-Weiner's, Simpson's, and Brillouin's) and similarity (Bray-Curtis). In addition, taxonomic sufficiency was determined for each of the metrics examined. For each of the analyses performed, both the statistical power of the analysis and the sample sizes required to achieve a desired level of power were determined. New methods are presented to deal with zero count data. Suggestions on how the power of microcosm tests used in toxicity testing can be increased are provided. The statistical programs developed for the analyses applied to the data analyzed were done in S-PLUS. The programs and directions for their application have been placed on an anonymous FTP site at the University of North Texas.