Development of geographically weighted regression using polynomial function approach and its application on life expectancy data

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)

Abstract

Geographically Weighted Regression (GWR) is a varying coefficient model. However, as an extension of Ordinary Linear Regression (OLR), it models a dependent variable at each location as a linear function of a set of independent variables. In real life, one or more independent variables involved in the model may have nonlinear relationships with the dependent variable. For this case, the GWR model is no longer realistic to use since the resulted analysis lead to be misleading. To overcome the problem, we develop the GWR by using a polynomial function approach. Here, the model is called Geographically Weighted Polynomial Regression (GWPolR). This paper aims to provide an algorithm, based on Akaike Information Criterion (AIC), for finding the optimal bandwidth and polynomial degrees. Furthermore, this paper aims to analyze life expectancy data in East Java province, Indonesia based on human development index and per capita expenditure. Compared with OLR and GWR models, GWPolR gave a significant improvement of goodness of fit measures and a more complete explanation of how each independent variable was related to the life expectancy.

Original languageEnglish
Pages (from-to)271-289
Number of pages19
JournalInternational Journal of Innovation, Creativity and Change
Volume5
Issue number3
Publication statusPublished - 2019

Keywords

  • Geographically weighted polynomial regression
  • Life expectancy
  • Spatial analysis

Fingerprint

Dive into the research topics of 'Development of geographically weighted regression using polynomial function approach and its application on life expectancy data'. Together they form a unique fingerprint.

Cite this