Analysis of the Poverty Level Model for West Sumatra Province Using Geographically Weighted Binary Logistic Regression

Abstract

Poverty is a widespread social problem that affects many developing countries, including Indonesia. The province of West Sumatra has a relatively low poverty rate of around 5.92 percent, making it the third lowest on the island of Sumatra. However, there are several districts and cities in this province that still have many people living in poverty. Various factors such as income levels, social conditions, and access to education, can contribute to the poverty gap in various regions. Geographically Weighted Binary Logistic Regression (GWBLR) is used to examine the relationship between poverty and geographic factors. GWBLR is a statistical analysis technique that takes geographic variables into account when the response variable is categorical or dichotomous. This approach incorporates a bandwidth-dependent weighting function. By conducting a fit test using R software, it is known that the Fcount value is greater than the Ftable value, indicating a significant difference between the logistic regression model and GWBLR. The results show that the GWBLR model with Fixed Gaussian Kernel weights is the most effective in analyzing poverty in the province. This model shows the lowest Akaike Information Criterion (AIC) value. Furthermore, this study identifies the Life Expectancy Variable as a significant factor affecting poverty in certain districts and cities in West Sumatra Province in 2022.