Analisis Transformasi Box Cox Untuk Mengatasi Heteroskedastisitas dalam Model Regresi Linier Sederhana

Abstract

Regression analysis is one of statistic technics that used to determine the relation model of one respon variable (Y) with one or more independent variable (X), what is generally expressed in equation mathematic. In statistic, a regression model is obtained by estimate of its parameter by using certain method, one of them is with the Maximum Likelihood Methods. Regression model that obtained to be told is good or fit, if fulfilled by the ideal assumption (classic). One of linear regression assumption which must be fulfilled is homogeneity varian (variant from error have the character of constant) so called also homoscedasticity. On the contrary, in reality if varian from error is not constant for example big or minimize higher at value X, so the condition told to heteroscedasticity or written down by: ar , , , . In regression model if all classic assumption were fulfilled, except one of them was the heteroscedasticity, so estimator that obtained still unbiased and consistent, but inefficient (big varian). One of way to overcome the heteroscedasticity in regression model is by Box Cox Transformation. Box Cox Transformation that is do the transformation to respon variable Y which be ranked with the parameter , so that become and estimator of parameter that obtained residing in gyration (-2,2).