Robust Path Analysis Development for Indonesian Economic Growth Determinants
Fitri Catur Lestari (a*), Anang Kurnia (b), Aji Hamim Wigena (b), Utami Dyah Syafitri (b)

a) Politeknik Statistika STIS
Jl. Otista 64C 13330 Jakarta, Indonesia
*fitricaturlestari[at]stis.ac.id
b) Institut Pertanian Bogor
Jl. Raya Dramaga Bogor 16680 West Java, Indonesia

Abstract

Outliers are disturbances in normality assumption of correlation and regression analysis. The classical path analysis that^s formed from them is also suspected to be prone to outliers. The purpose of this research are to develop a robust path analysis from classical one without eliminating outliers and to apply it in real data. The secondary data are collected from BPS Statistics Indonesia for 33 provinces. Eight variables (economic growth [X1], ratio Net Enrollment Rate [NER] of girls to boys in primary and high school education level [X2&X3], percentage ratio of female to male in the labor force [X4], percentage ratio of women to men who manage household [X5], mean years school [X6], urbanization [X7] and unemployment level [X8]) and five models are described in path diagram based on previous research. Robust path analysis development doesn^t need outlier eliminating and the result is better than the classical one. The variety of economic growth is explained 99.1% by model in robust path analyis and only 63.4% in classical one. Based on robust path analysis, X1 is respectively 0.13929, 0.07049, 0.03513 and 0.00186 directly affected by X2, X3, X4 and X5. The highest indirect impact is X7 to X1 which is 0.0199 through X3.

Keywords: economic growth, outliers, robust path analysis

Topic: MATHEMATICS AND STATISTICS