A rigorous, reproducible spatial data science framework engineered to investigate, model, and quantify the structural and socioeconomic drivers of regional life expectancy disparities across English Local Authorities.
This project investigates, models, and quantifies the structural and socioeconomic drivers of regional life expectancy disparities across English Local Authorities. By deploying a staged econometric modeling framework and spatial clustering architectures, the research maps and assesses the historic North-South health gradient, providing a statistically robust basis for identifying key predictors of regional health inequality.
The Main_Dataset.csv included in this repository was generated by merging five distinct official datasets from the Office for National Statistics (ONS) and the UK Government portal.
To replicate the spatial pipelines and modeling architectures locally:
- The
Main_Dataset.csvis provided here for immediate analysis. - The raw source datasets were obtained from:
- Life Expectancy Data: Office for National Statistics (ONS) - Local Areas.
- Indices of Deprivation (2019): UK Government Open Data Portal.
- Demographics & Boundaries: ONS Open Geography Portal.
- The
England_Life_Expectancy_Analysis.ipynbpipeline performs all necessary data wrangling, cleaning, and merging required to transform these raw sources into the final analytical dataset.
The analysis follows a professional data science pipeline, as detailed in the included research report:
- Data Preparation & Cleaning: Processed and unified multiple ONS and government datasets, mapping them to Local Authority District (LAD) codes and standardizing feature sets.
-
Feature Engineering: Applied logarithmic transformations (
$\log_{10}$ ) to stabilize right-skewed infrastructural variables (e.g.,Population_Density) and performed Z-score normalization for coefficient comparability. - Market Segmentation: Deployed an unsupervised K-Means clustering algorithm on life expectancy metrics to compute a mathematically objective threshold, establishing a data-driven baseline for regional health divides.
- Explanatory Modeling: Developed a series of 5 staged multivariable OLS regressions guided by a Directed Acyclic Graph (DAG) to isolate structural path mediators, controlling for heteroscedasticity using HC3 Robust Standard Errors.
-
Visual Analytics: Implemented spatial mapping and regression diagnostics using Python libraries to validate model explanatory power (
$Model\ 5\ \text{Adj.}\ R^2 = 0.914$ ).
In line with academic integrity and professional transparency, this project utilised AI assistance (Gemini) as a technical co-pilot:
- Troubleshooting: Used to resolve statistical diagnostic issues, including the application of HC3 Robust Standard Errors following Breusch-Pagan test failures.
- Optimisation: Used to improve the efficiency of statistical plotting loops and vectorising data manipulation logic for better performance.
- Verification: Used to cross-reference econometric definitions and ensure the final modeling pipeline addressed the requirements of the project mark scheme.
- Author's Role: All spatial interpretations, policy implications, and final model selections are the original work of the author.
- Python: Pandas, Numpy, Scipy, Statsmodels, Scikit-learn, Matplotlib, Seaborn.
- Licence: Apache License 2.0.
England_Life_Expectancy_Report.pdf: Full technical research report, economic context, and statistical validation.England_Life_Expectancy_Analysis.ipynb: Complete Python implementation with data cleaning, OLS regression models, and diagnostic plots.metadata/: Standardized Dublin Core schema definitions and explicitdc:creatordata governance records.