Using of Factor Analysis Scores in Multiple Linear Regression Model for Prediction of Kernel Weight in Ankara Walnuts

Abstract

Kernel weight is important for plant breeders to select high productive plants. The determination of relationships between kernel weight and some fruit-kernel characteristics may provide necessary information for plant breeders in selection programs. In the present study, the relationships between kernel weight (KW) and 7 fruit-kernel characteristics: Fruit Length, (FL,), Fruit Width (FW) Fruit Height (FH) Fruit Weight (FWe) Shell Thickness (ST), Kernel Ratio (KR) and Filled-firm Kernel Raito (FKR,), were examined by the combination of factor and multiple linear regression analyses. Firstly, factor analysis was used to reduce large number of explanatory variables, to remove multicolinearty problems and to simplify the complex relationships among fruit-kernel characteristics. Then, 3 factors having Eigen values greater than 1 were selected as independent or explanatory variables and 3 factor scores coefficients were used for multiple linear regression analysis. As a result, it was found that three factors formed by original variables had significant effects on kernel weight and these factors together have accounted for 85.9 % of variation in kernel weight.