Continum Regression with Discrete Wavelet Transformation Preprocessing

Abstract

In the multiple regression modeling, serious problems will be occurred if  independent variables are correlated, to be named  ill conditioned problem, and the number of observations is much less than the number of independent variables, it is a singularity problem. Continum Regression (CR) approach, it’s better to overcome the problem of ill conditioned, but if  the number of observations is much less than the number of independent variables usually facing the problem in computing. So the first step, it needes dimension reduction of independent variables (known as a preprocessing method). Discrete wavelet transformation (DWT) is one of a good method handle the problem of singularity.  The research we have studied  combination of CR and DWT as a preprocessing method can solved the problems of ill conditioned and singularity.  The result of empirical research with simulation data has concluded that performance of CRDWT have very good potency to overcome the problems of the number of observations much less than the number of independent variables and ill conditioned.