PROCEDURE


Market Research Center analysts were responsible for managing and manipulating the raw data sets to obtain the
required data subsets, i.e., general classes of seafood species and demographic classifications. Personnel in the
University of Florida's Department of Statistics then generated basic distributional statistics (means, medians and
percentiles) of per capital consumption for each general seafood classification in total, as well as for demographic
classifications. For each estimate provided, ninety-five percent confidence intervals were computed, provided sample
sizes were adequate. Because sample sizes for some demographic classifications were quite small, confidence intervals
for a number of the estimates were unobtainable. Estimates of percentiles were based on appropriately smoothed
empirical density functions (Silverman, 1986).

Differences in average per capital consumption among demographic classifications were analyzed by comparing their
respective empirical distribution functions using the Kolmogorov-Smirinov test (Hollander and Wolfe, 1973).
Computations were performed using SAS (SAS Institute Inc.) and S-plus (StatSci Inc.) statistical software.

Market Research Center personnel created and documented the electronic data sets for future use by HRS and DEP.