Metode Regresi Komponen Utama, Regresi Kuadrat Terkecil Parsial, dan LASSO pada Data Kemiskinan Hasil Olahan Susenas 2012

Abstract

Multicollinearity is a problem that is arise in multiple linear regression. Multicollinearity lead testing and estimating coefficient regression models become invalid because it produces a large variety. Methods that can be used to overcome multicollinearity in the data in among these are PCR, PLS, and LASSO. PCR and PLS forming independent new components to overcome multicollinearity. Both method previously unable to do the screening variables. LASSO does the screening variables by shrinking appropriate coefficient value of zero. The result showed the best model of the three based on RMSE value produced by LASSO method, and based on RMSEP value produced by RKU method. Difference in value of error for the three method are not much different.