Bayesian SIR Parameter Estimation (MH + Gibbs + Adaptive MCMC)

This repository provides an R implementation for Bayesian parameter estimation in a Susceptible–Infected–Removed (SIR)-type epidemic model using Metropolis–Hastings (MH) and Gibbs sampling.

Results

1️⃣ Standard Metropolis–Hastings Sampling

Each panel shows the parameter trajectory across iterations. The red line marks the true simulated value. This version uses a fixed proposal covariance, resulting in slower convergence.

2️⃣ Adaptive MH + Gibbs Sampling

Adaptive MH automatically updates the proposal covariance using the running chain covariance, leading to smoother mixing and faster convergence. Parameters stabilize around their true values more efficiently.