Open AccessFixing multicollinearity in modelling market body weight of Sudani ducks (Cairina moschata) from early age morphometric traits
Author(s) -
Abel Oguntunji,
A. Makram
Publication year - 2019
Publication title -
agricultural science and technology
Language(s) - English
Resource type - Journals
eISSN - 1314-412X
pISSN - 1313-8820
DOI - 10.15547/ast.2019.04.049
Subject(s) - multicollinearity , statistics , variance inflation factor , mathematics , collinearity , stepwise regression , regression analysis , coefficient of determination , linear regression , zoology , akaike information criterion , body weight , biology , veterinary medicine , medicine , endocrinology
Morphometric parameters and their indices are central to characterization, selection and genetic improvement of farm animals. The present study was conducted to fix multicollinearity among predictors and to fit optimum regression model for prediction of 14-week body weight of 150 un-sexed Sudani ducks from early age (0, 2, 4 and 6 weeks of age) morphometric measurements (body weight, body length, breast circumference and keel length). Pairwise correlation between the 14-week body weight (BWT) and measured variables was the highest with 2-week BWT (r=0.934, P<0.001). Application of multicollinearity diagnostics (variance inflation factor, tolerance, eigenvalues and condition index) revealed collinearity among six variables. Stepwise regression equation identified 2-week BWT as the most important predictor accounting for 87.30% of the total variation of the response variable. However, the optimum regression model for prediction of adult body weight was fitted with 2-week BWT in quadratic model having the highest predictive power (Coefficient of determination /R2/=0.966 and least root mean square error /RMSE/=4.269). This result is applicable under field conditions for both researchers and duck farmers for early selection of 14-week market body weight of Sudani ducks at 2 weeks of age.