Stepwise canonical discriminant analysis for morphometric characterization of three strains of broiler chicken

Abstract

Stepwise canonical discriminant analysis of quantitative traits was performed to characterize Arbor Acre (n = 29), Marshall (n = 29) and Ross 308 (n = 30) strains of broiler chicken at two weeks of age. Average body weight of 322.11 ? 51.33 g, 310.70 ? 51.20 g and 310.24 ? 42.37 g were obtained for Arbor Acre, Marshall and Ross 308 strains, respectively. Correlation coefficients between traits was positive (0.13-0.80). Out of the eight traits used, shank length (SL), body length (BL), wing length (WL), breast girth (BG) and thigh circumference (TC) were selected by stepwise procedure as discriminating variables with respective tolerance of 0.75, 0.84, 0.86, 0.57 and 0.66. The Mahalanobis distance indicated that Arbor Acre and Marshall were closely related by SL (2.006), WL (5.704), BG (6.002) and TC (6.314) while Arbor Acre and Ross 308 were related by BL (4.212). Two canonical discriminant functions were generated. Function 1, with smaller Wilk's Lambda (0.15), stronger canonical correlation coefficient (0.88), highly significant Chi-square (p<0.0001), greater eigenvalue (3.14) and percentage variance (87.28%), possessed higher discriminatory power than function 2.72.4%, 82.8% and 90.0% of the cross validated cases were correctly classified as Arbor Acre, Marshall and Ross broilers, respectively. The study revealed that Ross 308 was genetically distinct from Arbor Acre and Marshall, which were more closely related, using shank length, body length, wing length, breast girth and thigh circumference as discriminating traits. Stepwise canonical discriminant analysis is therefore relevant in classification genetics.