Comparison of Ridge Regression and Principal Component Regression in Overcoming Multicollinearity of Factors Affecting Poverty in Indonesia

Regression analysis is a statistical method that aims to measure the correlation or relationship between the independent variables on the dependent variable. The Regression analysis has several assumptions that must be met to obtain a good model. One of the assumptions that must be fulfilled is the absence of multicollinearity between the independent variables. Multicollinearity is a situation where there is a relationship or correlation between independent variables which causes the accuracy of the model to decrease. This study aims to compare ridge regression and Principal Component Regression (PCR) in overcoming multicollinearity. Ridge regression is a modification of the least squares method that adds a bias constant to the main diagonal of the variance–covariance matrix. The calculation of the bias constant in this study uses the Kibria method. Principal component regression is a combination of Principal Component Analysis (PCA) and regression analysis, this method works by regressing the main components. The comparison criteria used are \({R}^{2}\) and \({R}_{adj}^{2}\). The research data used is on factors influencing poverty in Indonesia in 2021. Based on the calculation results, it is found that the accuracy of the ridge regression model is 83.6%, while the accuracy of the principal component regression model is 77.5%, this value is seen from the largest value of \({R}^{2}\). Furthermore, the \({R}_{adj}^{2}\) value of the ridge regression model is greater than the \({R}_{adj}^{2}\) value of the principal component regression model, namely 0.7991 > 0.761. Thus, it can be concluded that ridge regression with Kibria parameters is better than principal component regression in overcoming multicollinearity in data on factors that influence poverty in Indonesia.