International Conference on Multidisciplinary, Engineering, Science, Education and Technology (IMESET’17 Bitlis), Siirt, Turkey, 27 - 29 October 2017, pp.290
In this study, egg albumen weight has determined by using egg external quality components and variable selection has been made. In line with this purpose, Ridge and LASSO regression analysis techniques known as penalized methods are used. The dataset contains the internal and external quality of 117 Japanese quail eggs. The independent variables in the model are the external quality components; egg width, egg size, egg weight, shape index and eggshell weight. According to the results obtained by using the OLS method; Mean Squared Error (MSE) and coefficient of determination (R2) are 0.04987 and 0.7699, respectively; the VIF values for egg width, egg size and shape index were 827, 416 and 1197, respectively. VIF ≥ 10 indicates that there is a multicollinearity problem. In order to overcome this problem, the Ridge and LASSO regression analysis were applied. The variable selection is not performed in Ridge regression. In the model with five variables the Ridge regression analysis results are revealed that MSE = 0.05009 and R2 = 0.7693 for Ridge parameter 0.015. In LASSO regression, estimation and variable selection have been made at the same time and two variables have been selected; egg weight and eggshell weight. According to LASSO regression analysis results, MSE and R2 values have been obtained as 0.04981 and 0.7638 respectively. LASSO is obtained very close to the coefficient of determination that the ridge regression is possessed by only two variables. For data sets with multicollinearity problems, the LASSO regression should be used instead of the ridge regression if the goal is not only estimation but also variable selection. When LASSO regression, which is a biased estimation technique similar to ridge regression, is used, more consistent and reliable estimations are obtained compared to the OLS method and variable selection is also made at the same time.