Aplikasi generalized ridge regression dalam mengatasi autokorelasi dan multikoliniaritas pada INDesKS harga saham gabungan di Bank Indonesia

INDONESIA:

Analisis regresi linier berganda terkadang terdapat masalah autokorelasi dan multikolinieritas. Kombinasi antara Generalized Ridge Regression dan Two Stage Least Square adalah salah satu metode yang digunakan untuk mengatasi masalah autokorelasi dan multikolinieritas. Dalam penelitian ini kombinasi antara Two Stage Least Square dan Generalized Ridge Regression disebut Two Stage Generalized Ridge Regression. Tujuan penelitian ini yaitu: (1) penggunaan Two Stage Least Square dan Generalized Ridge Regression untuk mengatasi masalah autokorelasi dan multikolinieritas, (2) perbandingan antara estimasi Two Stage Generalized Ridge Regression dan estimasi Ordinary Least Square.

Hasil Analisis menunjukkan bahwa metode Ordinary Least Square mengalami masalah autokorelasi positif karena nilai Durbin Watson 1,345<d_L 1,494 serta mengalami masalah multikolinieritas karena nilai VIF peubah bebas kurs USD dan IDJ di atas 5. Penelitian ini ada dua tahap yaitu: 1) mendeteksi data yang mengalami masalah autokorelasi dan multikolinieritas menggunakan metode Ordinary Least Square dan analisis Two Stage Least Square, 2) estimasi Generalized Ridge Regression yaitu mentransformasi data menggunakan metode centering dan rescaling, menentukan nilai k, menentukan persamaan Two Stage Generalized Ridge Regression, dan transformasi persamaan Two Stage Generalized Ridge Regression.

Hasil analisis menggunakan metode Two Stage Generalized Ridge Regression yaitu nilai Durbin Watson 4-d>d_U dan nilai VIF setiap peubah bebas di bawah 5, sehingga masalah autokorelasi dan multikoliniertas telah diatasi. Bentuk persamaan pada pengaruh suku bunga Sertifikat Bank Indonesia, kurs USD, dan Indeks Dow Jones terhadap Indeks Harga Saham Gabungan menggunakan metode Two Stage Generalized Ridge Regression adalah
IHSG=-116,765+3,975 SBI +0,045 USD+6,270 INF+0,036 IDJ

ENGLISH:

Multiple linear regression analysis sometimes has autocorrelation and multicollinearity problem. The combination between Generalized Ridge Regression and Two Stage Least Square are the method to overcome the problem of autocorrelation and multicollinearity. In this research the combination of Two Stage Least Square and Generalized Ridge Regression are called Two Stage Generalized Ridge Regression. The purpose in this research are (1) applying Two Stage Least Square and Generalized Ridge Regression to overcome the problem of autocorrelation and multicollinearity, (2) comparing between Two Stage Generalized Ridge Regression estimation and Ordinary Least Square estimation.

The result of analysis indicate that Ordinary Least Square method has positive autocorrelation problem and multicolinearity problem because the value of Durbin Watson is 1,345 < d_L 1,494 and the value of VIF in the perdictor variable rate of USD excange and IDJ are above 5. This research has two stages as follows 1) applying Ordinary Least Square method and Two Stage Least Square analysis to detect the data that have experience problem of autocorrelation and multicolinearity problem, 2) performing Generalized Ridge Regression estimation as follows: transform the data using centering and rescaling methods, determining the k value, determining the Two Stage Generalized Ridge Regression equation, and transforming the Two Stage Generalized Ridge Regression equation.

The result analysis using Two Stage Generalized Ridge Regression method shows that the value of Durbin Watson is 4-d> d_U and the VIF value of every independent variable is under 5, so the problem of autocorrelation and multicolinearity are resolved. The form of regression equation on the influence rate of Bank Indonesia Certificates, USD exchange rate and Index Dow Jones against Composite Stock Price Index using Two Stage Generalized Ridge Regression method as follows
IHSG=-116,765+3,975 SBI +0,045 USD+6,270 INF+0,036 IDJ