Categorization of seafood quality using discriminant function analysis over several decomposition variables
A multivariate statistical procedure, discriminant function analysis (DFA), was used to categorize seafood samples into one of three quality classes: fresh, marginal or unacceptable. Controlled time-decomposition studies helped establish spoilage patterns for preclassification of individual samples for DFA. The quality of lean fish, fatty fish and shrimp were predicted by DFA in three separate studies using sensory analysis and decomposition data from multiple spoilage indices. Prediction accuracy was determined using all test indices and by a systematic computer elimination of non-significant spoilage tests at $p<0.05.$^ The statistical analysis correctly classified 98.5% of the lean fish samples (n = 67), 86.2% of the fatty fish samples (n = 58) and 98.7% of the shrimp samples (n = 79) using all the quality indices. Computer selection of predictor indices yielded correct classifications of 95.5%, 81.0% and 97.5% respectively. The number of tests required to effectively categorize quality were reduced from fifteen to three for lean fish, thirteen to three for fatty fish, and eleven to six for shrimp, with minimal losses in prediction accuracy and a substantial reduction in analysis time. ^
Agriculture, Food Science and Technology|Health Sciences, Public Health
Paul Christopher Ellis,
"Categorization of seafood quality using discriminant function analysis over several decomposition variables"
Dissertations and Master's Theses (Campus Access).