PERBANDINGAN METODE ROBUST MCD-LMS, MCD-LTS, MVE-LMS, DAN MVE-LTS DALAM ANALISIS REGRESI KOMPONEN UTAMA

Abstract

Principal Component Regression (PCR) is one of the widely used statistical techniques for
regression analysis with colinearity. A robust technique on CR required is when data contains
outlier is urgently needed.
In this research we consider combination between Robust Principal Ccomponent Analysis
(PCA): Minimum Covariance Determinant (MCD) and Minimum Volume Ellipsoid (MVE) with
Robust Regression methods: Least Median Square (LMS), and Least Trimmed Square (LTS), then
compare resistance level of MCD-LMS, MCD-LTS, MVE-LMS and MVE-LTS through the bias
and the mean square error on some samples size and outlier’s percentage.
The result shows that the MCD-LMS perform better than MCD-LTS, MVE-LMS, and MCDLTS.