PENDEKATAN MIXED GEOGRAPHICALLY WEIGHTED REGRESSION UNTUK PEMODELAN PERTUMBUHAN EKONOMI MENURUT KABUPATEN/KOTA DI JAWA TENGAH

WIDAYAKA, PRATAMA GANANG (2016) PENDEKATAN MIXED GEOGRAPHICALLY WEIGHTED REGRESSION UNTUK PEMODELAN PERTUMBUHAN EKONOMI MENURUT KABUPATEN/KOTA DI JAWA TENGAH. Undergraduate thesis, Fakultas Sains dan Matematika, Undip.

[img]
Preview
PDF
2269Kb

Abstract

Global regression models with a multitude of residual variance in each region causing non-homoskedastisitas assumptions are not met. The diversity of the geographic location factors causing spatial heterogeneity. Geographically Weighted Regression (GWR) is a development of linear regression by involving diverse factors geographical location, so that the parameters generated will be local. GWR model is not able to model the combination of local and global influences in a model. So the purpose of forming a GWR Mixed models are able to establish a model GWR with local and global influences simultaneously. GWR Mixed Model is used to estimate the model Gross Regional Domestic Product (GRDP). As independent variables that influence is revenue (PAD / X1), a variable amount of labor (JAK / X2), the human development index (HDI / X3), unemployment rate (TPT / X4) and the regional minimum wage (UMR / X5 ). Mixed GWR model the variables that are local and which are global variables. Methods for estimating model parameters MGWR using Weighted Least Square (WLS). Weights obtained the appropriate model to estimate the optimal bandwidth by using the reference method Cross Validation (CV) is a minumum. MGWR models with adaptive exponential kernel function weighting on Gross Domestic Product in the districts / cities in Central Java to produce variable JAK, IPM and TPT have the nature of the locality an area that is significant to the later model PAD have a global nature that sigbifikan against the model. To mengengetahui error rate value model is used Akaike Information Criterion (AIC). Keywords: Akaike Information Criterion, Bandwidth Cross Validation, Fungsi Kernel Gaussian, Mixed Geographically Weighted Regression, Weighted Least Square.

Item Type:Thesis (Undergraduate)
Subjects:H Social Sciences > HA Statistics
Divisions:Faculty of Science and Mathematics > Department of Statistics
ID Code:55041
Deposited By:Mr Hasbi Yasin
Deposited On:25 Jul 2017 14:11
Last Modified:25 Jul 2017 14:11

Repository Staff Only: item control page