PERBANDINGAN REGRESI RIDGE DAN PRINCIPAL COMPONENT ANALYSIS DALAM MENGATASI MASALAH MULTIKOLINEARITAS

Abstract

Multiple linear regression said to be good if it statistic the
assumptions such as: normality assumption, heteroskedastisity, an error
does not undergo autocorrelation and not occour multicolinearity. On the
assumption that problem often arise in the multiple linear regression
assumptions are not fulfilled multicolinearity. Multicollinearity is a
condition in which the data of the observations of the independent variables
occuror have a relationship that is likely to be high. This study aimed to
compare the appropriate method to over come multicollinearity between
ridge regression and principal component analysis. Comparison criteria
used both methods, the mean square error (MSE) and the coefficient of
determination (R2), from the data is the simulation with Microsoft Excel
then the analysis was performed, in order to obtain the data first using ridge
regression has a value of MSE of 0.02405 and R2 of 82.4%, while the
principal component analysis MSE value of 14.14 and R2of 37.5% while the
data second using ridge regression MSE has a value of 0.00216 and R2 of
96.9%, while the principal component analysis MSE values of 5.15 and R2 of
69.5%. From these results it can be concluded that ridge regression method
is better used.