Trakia Journal of Sciences · 2009 · 47 citations · 14 references
Open access
Life SciencesAnimal SciencesZoologyAnimal AnatomyMammalogyRuminant MedicineBiometricsBody CompositionStature EstimationBivariate CorrelationBody Size EstimationCorrelationCorrelation MatrixCorrelation Matrix EstimationGrowth Trait CorrelationsStepwise Regression
Morphobiometric studies were conducted on 58 West African dwarf kid goats of both sexes in order to determine the problem of collinearity in body weight estimation. There were no gender-associated differences in the estimated morphological indices. The bivariate correlations between body weight and zoometrical variables were positive and highly significant. The highest correlation was observed between withers height and rump height while the lowest was recorded for withers height and rump width. Variance inflation factors (VIFs) showed severe collinearity problems in three body measurements as these were higher than 10.00 (VIF= 24.992, 22.377 and 12.074 for withers height, rump height and body length respectively). This was confirmed by tolerance (T) values lower than 0.1 in the same measurements. Eigenvalues of the correlation matrix, condition indexes and variance proportions showed also collinearity. Among the collinear variables, withers height was retained as the most single important trait for prediction. The eventual stepwise multiple regression models revealed that body weight was best predicted from a combination of withers height, heart girth and rump width [R 2 , Adjusted R 2 and RMSE (Root mean squares error) = 0.951, 0.949 and 0.698 respectively].
14