Andoh-Omotayo/bayesian-amr-nigeria

Bayesian hierarchical logistic regression analysis of antimicrobial resistance (Ciprofloxacin) using PyMC and MCMC methods. # Bayesian Hierarchical Modeling of Antimicrobial Resistance in Nigeria ## Overview This project applies Bayesian hierarchical logistic regression using PyMC to analyze antimicrobial resistance (AMR) patterns for Ciprofloxacin. The goal is to estimate the probability of resistance while accounting for: * Multidrug resistance (MDR) * Gene detection status * Specimen-level variation * Age-group variation This approach provides full uncertainty quantification through posterior distributions using Markov Chain Monte Carlo (MCMC) sampling. ## Why Bayesian? Traditional logistic regression gives single-point estimates. Bayesian modeling provides: * Probability distributions for each parameter * Credible intervals (HDI) * Full uncertainty quantification * Better modeling of hierarchical biological structure This is critical in AMR research where uncertainty directly affects treatment policy decisions. ## Model Structure Outcome: * Ciprofloxacin resistance (Binary) Fixed effects: * MDR status * Gene detection Random effects: * Specimen group * Age group Link function: * Logistic (sigmoid) Inference method: * No-U-Turn Sampler (NUTS) ## Key Results * MDR strongly increases probability of resistance. * Gene presence shows weaker but positive association. * Variability exists across specimen types and age groups. * All chains converged (R-hat ≈ 1.00). ## Tools Used * Python * PyMC * ArviZ * NumPy * Pandas Author Omotayo Andoh Statistics | Bayesian Modeling | AMR Research