Estimation of egg albumen height in Nick Brown hens: A comparison of Ridge and LASSO regression methods

This study aims to estimate albumen height in eggs obtained from Nick Brown hens using external egg quality traits. Egg weight, egg width, egg length, shape index, Haugh unit, shell weight, and shell thickness are included in the model as independent variables. Multiple linear regression analysis is first applied, and the model is statistically significant (R² = 0.94, r = 0.97). However, strong relationships among variables and high VIF values indicate that model coefficients may be unstable and that multicollinearity may affect the results. Ridge and least absolute shrinkage and selection operator (LASSO) regression methods are therefore applied to obtain more reliable and stable estimates. These approaches are evaluated to improve model stability and enhance prediction performance in the presence of highly correlated variables. The findings indicate that LASSO regression demonstrates better performance than Ridge regression, with higher explanatory power (R² = 0.95) and lower information criteria (AIC, AICc, and BIC). In conclusion, penalized regression methods provide more reliable estimates than multiple linear regression in data structures with multicollinearity, and LASSO regression is more effective in terms of model performance and model parsimony.