PERBANDINGAN REGRESI KOMPONEN UTAMA DENGAN REGRESI KUADRAT TERKECIL PARSIAL PADA INDEKS PEMBANGUNAN MANUSIA PROVINSI JAWA TIMUR

*Vetranella .T.R.A. Sinaga  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Diah Safitri  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Rita Rahmawati  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Published: 29 Nov 2019.
Open Access Copyright 2020 Jurnal Gaussian
License URL: http://creativecommons.org/licenses/by-nc-sa/4.0

Citation Format:
Abstract

The multiple regression classic assumptions are used to give linear unbiased and minimum variance estimator. In Human Development Index (HDI) and influencing factors in East Java, there are two variables with VIF more than 10 so the assumption of non-multicollinearity is not fulfilled. Principal component regression (PCR) and partial least squares regression (PLS-R) can solve this problem. By doing principal component analysis, there are two linear combinations to take as the new   independent variables which are free from collinearity. In the PLS-R, NIPALS algorithm is used to calculate the components and other structures and to estimate the parameter. While in PCR all independent variables are significant, the percentage of households with drinking water is feasibles is not significant in the model. PLS-R’s  is 95,85% is greater than PCR’s  = 93,42%. PCR’s PRESS = 81,78 is greater than PLS-R’s PRESS = 61,0595.

Keywords: Human Development Index (HDI), Multicollinearity, Principal Component Regression, Partial Least Squares Regression, , PRESS

Keywords: Human Development Index (HDI), Multicollinearity, Principal Component Regression, Partial Least Squares Regression, , PRESS

Article Metrics: