Correction for multicollinearity between the explanatory variables to estimation by using the Principal component method

Abstract

In the most of applications of regession ,no explanatory orthogonal will exist,but it is connected too strongly to the extennt that the results would be far from being exquisite. So its too difficult to expect the effects upon the individual variabls within the range of regression equation .Also the values estimated here concerning the factors could be slight in data. The non-orthogonality is said to be the problem of multicolinearity going side by side with the factors of unstable,estimated regression. This case,however,comes out from the strong linear relation between explanatory variants.To solve this problem, a method of the pricipal components is used;that which depends upon the fact that each linear type mihgt be reformulated as to the group of the orthogonal,explanatory variables; these in turn can be obtained as linearstructures for the orthogonal (basic) explanatory variables through the Barr'let norm,as a test formula, as away to know whetherthe roots possess a sufficient quality for a linear relation As for the practical side,or applied of this research concerning the special
data of the consumption of the individual in USA as dependent variable and wage income,non wage-non farm income,farm income are explanatory variables.the characters rootindicated to the collinearity ; the result is that the four variables can be treated as two factors only ,a ststistical programme is here used that is (SPSS,Minitab) for the analysis of the data.