PENERAPAN REGRESI AKAR LATEN DALAM MENANGANI MULTIKOLINEARITAS PADA MODEL REGRESI LINIER BERGANDA

Abstract

Multicollinearity is a problem that often occurs in multiple linear regression. The existence of multicollinearity in the independent variables resulted in a regression model obtained is far from accurate. Latent root regression is an alternative in dealing with the presence of multicollinearity in multiple linear regression. In the latent root regression, multicollinearity was overcome by reducing the original variables into new variables through principal component analysis techniques. In this regression the estimation of parameters is modified least squares method. In this study, the data used are eleven groups of simulated data with varying number of independent variables. Based on the VIF value and the value of correlation, latent root regression is capable of handling multicollinearity completely. On the other hand, a regression model that was obtained by latent root regression has   value of 0.99, which indicates that the independent variables can explain the diversity of the response variables accurately.