Classifying five ornamental fish species of Cichlidae family by use of logistic regression and discrimination linear analysis

Abstract

This experiment was conducted to classify five ornamental fish species, Heros severus, Astronotus ocellatus, Pterophyllum scalare, Amphilophus citrinellus × Paraneetroplus melanurus, and Amphiprion perideraion, regarding the growth indices BWI, SGRW, and SGRL by use of Logistic Regression and Discrimination Linear Analysis, in a completely random design with five treatments (5 species), each with four replicates with a density of 25 specimens per replicate and totally equal to 500 ornamental fish. Feeding trial with iso-nitrogenous and iso-caloric diets lasted for 90 days, and every thirty days fish biometry was practiced. At the end of the trial, data were collected to find the best class for each species regarding the least distances with other species. The results obtained from two statistical models were compared and it was concluded that the Logistic Regression model in species classification based on the type of nutrition and considering the BWI and SGRW indices rather than the descriptions by Discrimination Linear Analysis, is more appropriate. However, the SGRL index was not considered as an appropriate factor for the separation of fish samples according to species factor in either of these two models.