Low Severity
Published: Oct 25, 2026
Undergraduate thesis examining environmental and socioeconomic predictors of climate vulnerability across 17 administrative regions of the Philippines (2010–2024) using a complementary Two-Way Fixed Effects panel regression and XGBoost machine learning framework.
# Environmental and Socioeconomic Predictors of Climate Vulnerability in the Philippines: A Complementary Panel Regression and XGBoost Analysis
## Overview
This repository contains the official dataset, statistical scripts, and empirical documentation for our undergraduate thesis completed in April 2026 at the Polytechnic University of the Philippines. This study investigates whether lower-emission regions bear disproportionately greater climate vulnerability, uncovering critical empirical evidence of a climate justice paradox in the Philippines.
---
## Key Visualizations
### Regional Climate Vulnerability in the Philippines
*Figure: Pooled Subnational GDL Vulnerability Index (SGVI) by Region (2010–2024).*
### Empirical Methodological Framework
*Figure: Proposed empirical workflow spanning multilevel data imputation, dimensionality reduction, specification testing, panel regression, and XGBoost machine learning benchmarking.*
---
## Methodology
This study utilizes a quantitative, non-experimental observational panel framework spanning all **17 administrative regions** of the Philippines from **2010 to 2024**.
1. **Data Reconstruction & Validation:**
* Addressed missing values using **Multiple Imputation by Chained Equations (MICE)** combined with **Two-Level Predictive Mean Matching (2L.PMM)** across 20 imputed datasets to preserve regional hierarchy and temporal continuity.
2. **Dimensionality Reduction:**
* Applied **Principal Component Analysis (PCA)** to construct composite indices: the *Socioeconomic-Environmental Quality Index (SEQI)* and the *Inequality Index*.
3. **Econometric Panel Regression:**
* Implemented a **Two-Way Fixed Effects (FE)** model equipped with **Driscoll-Kraay robust standard errors** to control for unobserved regional heterogeneity, serial correlation, heteros