ANALISIS KOMPONEN UTAMA DENGAN MENGGUNAKAN MATRIK VARIAN KOVARIAN YANG ROBUST

Abstract

The present of outlier data causes the estimator of variance-covariance be overestimated. As a consequent, in the principle component analysis, the variability of the data in the main component becomes biger than as expected. To cope this condition, one can use robust estimator, i.e. MVE and MCD. Using simulation of Monte Carlo Experiments, the Principal Component Analysis  using estimator MCD has the better performance than the estimator MVE and also classic estimator.