抄録
Regression analysis is the most classical but still a very useful back analysis in social sciences. Multicollinearity is one of the most serious ill-posedness and many conventional regularization methods have been provided. The paper focuses on the multicollinearity problem and attempts to compare the theoretical and practical characteristics among regularization methods. The paper discuses principal component regression and ridge regression which are the most well-known ones among them. It is shown that they are compatible with the traditional regularization methods in inverse analysis. The two methods are applied to regression analysis of land price in order to demonstrate their practical characteristics.
