EXPLORING DATA STRUCTURE OF NEW-YORK BIGHT BENTHIC DATA USING POST-COLLECTION STRATIFICATION OF SAMPLES, AND LINEAR DISCRIMINANT-ANALYSIS FOR SPECIES COMPOSITION COMPARISONS

Techniques for exploring data structure from four cruises on a rectilinear sampling grid system on the New York Bight are presented. The problem of microenvironmental variability is handled by post-collection stratification of samples based on the physical characteristics of each grab, rather than classical spatial strata definitions. The three physical variables used to define strata are: mean particle size, per cent organic matter, and total concentration of selected heavy metals. Variations in species densities within strata over time are investigated. Multivariate discriminant techniques are used to search for major statistical differences in a species abundance between strata. A flexible sampling program based on post-collection stratification is suggested to detect shifts in mean species abundance within strata over time. © 1979.