Statistical classification of seafood quality.

Discriminant function analysis (DFA) was used to classify the freshness quality of lean fish, fatty fish, and shrimp as either acceptable (Class 1), marginal (Class 2), or unacceptable (Class 3). Fresh and frozen survey samples were statistically classified following an initial precategorization using sensory, chemical, and microbiological indices as predictor variables. Computer elimination of nonsignificant predictor tests, p > 0.05, was used to optimize the test protocol. DFA correctly classified 98.5% of 67 preclassed lean fish samples (34 Class 1; 13 Class 2; 20 Class 3), 86.2% of 58 preclassed fatty fish samples (22 Class 1; 16 Class 2; 20 Class 3), and 98.7% of 79 preclassed shrimp samples (45 Class 1; 18 Class 2; 16 Class 3) by using all the quality indices. Computer selection of significant predictor indices at p < 0.05 yielded correct predicted classifications of 95.5, 81.0, and 97.5%, respectively. The number of tests required to effectively categorize quality were reduced from 15 to 3 for lean fish, from 13 to 3 for fatty fish, and from 11 to 6 for shrimp, with minimal losses in prediction accuracy and a substantial reduction in analysis time.