Distinguishing Beaked Redfishes (Deepwater Redfish, Sebastes mentella and Labrador Redfish, S . fasciatus ) by Discriminant Analysis (with Covariance) and Multivariate Analysis of Covariance

• Published in: Canadian Journal of Fisheries and Aquatic Sciences
• Main Authors: Misra, R. K., Ni, I-H.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Canadian Science Publishing 1983
• Subjects: Aquatic Science; Ecology, Evolution, Behavior and Systematics; Sebastes mentella
• Online Access: http://dx.doi.org/10.1139/f83-172; http://www.nrcresearchpress.com/doi/pdf/10.1139/f83-172

Analysis of morphometric data for differentiating fish species and stocks has frequently been unsatisfactory as a result of sampling bias associated with the varying size of specimens and the large overlapping of characters. These difficulties may be overcome by using a discriminant function with covariance and multivariate analysis of covariance. In a classification study of beaked redfishes, in which the specimens of Labrador redfish (Sebastes fasciatus) are relatively smaller than those of deepwater redfish (S. mentella), a single character would not separate species, but a compound criterion (discriminant function) of several characters separated the species effectively. A discriminant function with covariance separated species/populations remarkably well and better than one without covariance. Seven morphometric characters were identified as pertinent discriminators between S. fasciatus and S. mentella. As much as 89% of the total variation in the sample was accounted for by the discriminant function and only 8 out of 198 individuals (4%) were in the zone of uncertainty. It is explained why using a large number of characters in discriminant analysis may not be appropriate for samples of limited sizes. Also, expressing morphometric measurements as ratios, proportions, or percentages of body length may not be an appropriate way of reducing variation owing to size differences.