The Role of Level Selection as a Reference on Free Variable on Categorial Type on Multicolinierity Degree in Linier Regression Model.

Abstract

Utilizing free variable of categorial type in linier regression model tends to cause multicolinierity problems appear. Multicolinierity is also one factor causes error standard tobe so the interval trust for parameter is also bigger. The statis tic used to detect the indication of multicolinierity is condition number. The value of condition number depends on level selection category made as a reference of free variable of categorial type. The free variables are practically represented by dummy variables. There are 96 scenarios of change of categorial level as the reference. The result of scenario 1 is with eigen value of 6.344, 2.039, 1.309, 1.013, 1.012, 0.845, 0.669, 0.386, 0.142, 0.085, 0.073, 0.039, 0.029, dan 0.016. so form the calculation, it obtained that the value of condition number is 20.128, it means that condition number is in between 10-30 means that there is medium multicolinieritay problem. And the biggest value of condition number is in scenario of 78 with the value of condition number 51.242 whose value is in >30, this indicates that multicolinierity problem is considered serious. And the lowest value of condition number is in scenario of 50 and it is 17.891 and the value is between the score of10-30, it shows that there is a medium multicolinierity problem. The results show that multicolinierity problem in this research is medium with the distance score R-Square is 0.89.