Enhancing regression models: A comparative investigation of classical and ridge regression in the presence of multicollinearity in high density apple data

Abstract

The application of statistical principles and methods is necessary for effective practice in resolving the different problems that arise in many branches of agricultural activity. To find out one such suitable estimation procedures for Gala species of high density apple, the present investigation has been carried out. The study revealed that ridge regression produces coefficients with minimum mean square error and produces regression coefficients which predict better than ordinary least squares when predictor variables are highly correlated. The value of ridge constant (θ) was estimated by three different methods, viz., ridge trace technique, method given by Hoerl, Kennard and Baldwin and cross-validation method. Based on the minimum value of mean square error, ridge trace method was selected to calculate the optimum value of ridge constant.