FIXING MULTICOLLINEARITY INSTABILITY IN THE PREDICTION OF BODY WEIGHT FROM MORPHOMETRIC TRAITS OF WHITE FULANI COWS

Autor: Abdulmojeed YAKUBU
Sprache: Bulgarisch; Tschechisch; Englisch; Kroatisch; Ungarisch; Polnisch; Slowakisch
Veröffentlicht: 2011
Quelle: Directory of Open Access Journals: DOAJ Articles
Online Zugang: http://jcea.agr.hr/articles/74935_FIXING_MULTICOLLINEARITY_INSTABILITY_IN_THE_PREDICTION_OF_BODY_WEIGHT_FROM_MORPHOMETRIC_TRAITS_OF_WHITE_FULANI_COWS_en.pdf
https://doaj.org/toc/1332-9049
1332-9049
https://doaj.org/article/cd65de813a8947af819063fe5a817538
https://doaj.org/article/cd65de813a8947af819063fe5a817538
Erfassungsnummer: ftdoajarticles:oai:doaj.org/article:cd65de813a8947af819063fe5a817538

Zusammenfassung

Body weight and nine morphostructural characters (withers height, rump height, heart girth, body length, head width, cannon circumference, shoulder width, rump width and rump length) of 83 White Fulani cows aged 1.5-2.4 years old were used to study the problem of multicollinearity instability in the estimation of body weight from morphological indices. Pairwise phenotypic correlations indicated a high and positive significant relationship between body weight and body dimensions (r = 0.61- 0.94; P<0.01). Among the linear type traits, the highest correlation was observed between withers height and rump height (r =0.98) while the lowest value was recorded for rump height and shoulder width (r =0.51). Severe collinearity problems were evident in 5 of the zoometrical variables as portrayed by variance inflation factors (VIFs) higher than 10.00 (VIF = 33.096, 31.421, 24.612, 22.726 and 13.327 for rump height, withers height, rump length, heart girth and body length respectively). Collinearity problems were further confirmed from the computations of the eigenvalues of the correlation matrix, condition indexes and variance proportions. Heart girth was retained among the collinear variables, and singly accounted for 87.9%, 92.3% and 94.1% of the variation in body weight in the subsequent stepwise regression, quadratic and cubic models, respectively.