©️ Tags

Bayesian inference, Monte Carlo, logistic regression, classification

💡 About

The goal of the project is to discover factors leading to diabetes and to verify the prior assumptions based on the previous papers by relying on Bayesian inference.
The research is based on a dataset prepared by the National Institute of Diabetes and Kidney Diseases in the United States.

📂 Content

• data – folder with raw and cleaned datasets as well as .rstan files with prior features' distributions.
• data_preparation.R – initial data preparation procedure stored as R script
• analysis.R – analysis, modeling and testing procedure stored as R script
• Research.pdf – main file containing full research description

🧪 Methodology

Final estimates were achieved by applying the Hamiltonian Monte Carlo algorithm, which combined prior assumptions and estimations from the logistic regression model.

🏆 Findings

• Variables indicating the number of pregnancies, glucose, and BMI levels, as well as the age of the respondent met a priori assumptions 🎯
• The diabetes_pedigree_func variable, i.e. the variable responsible for the patient's genealogy, had a higher expected value than it was assumed a priori ❌
• The model also suggests that diabetes_pedigree_func is the most important factor: the greater the respondent's proximity in time and kinship to the last case of diabetes in the family, the greater the probability that the respondent will also suffer from diabetes ❤️‍🩹