다중공선성에 관한 연구 : Based on Principal Component Regression and Factor Analysis = A Study on Multicollinearity in Regression Diagonistics

다국어 초록 (Multilingual Abstract)

Regression diagnostics in the method of diagnosis accomplished for get more useful information in estimation of regression coefficient. And It diagnose investigation of error term hypothesis and correlation of independent variable, selection of fitted model, appearance of outlier. A value of measurement influenced effect by estimation of regression coefficient.
If there is mighty correlation among independent variable than It is difficult to get converse matrix of X'X or The mean of regression analysis is decline because of variance of regression coefficient estimated is expand and there is no meaning parameter estimation of method of least squares.
Multicollinearity is the case that it is difficult to get converse matrix of X'X or not exist as for being relation of mighty linear combination among of independent variable, each column of matrix.
Therefore, In this thesis, there is a target to reduce multicollinearity through the method of principal component regression and factor analysis in method of solution and the way of diagnose multicollnearity like this.
The method of principal component regression is the method of solve a problem that use principal component regression estimator was independent variable by principal component, in case appearance of multicollinearity.
In case, data of complicated from are correlated, we will interpret for search a potential common factor that can explain correlation among variables throughout factor analysis.
Solve the problem throughout the factor analysis, find potential common factor of complicate variables and establish of statistic model and then explain relation structure of variables and estimate factor score and that apply regression analysis.