Usage Guide
Bayesian NMA
How to run Bayesian network meta-analysis
1
Create a project and upload data
- 1Same as Frequentist NMA — create project, add outcome, upload data
- 2Data format is identical (arm-level or contrast-level)
2
Configure MCMC settings
- 1Set the number of MCMC iterations (default: 50,000)
- 2Set burn-in period (default: 10,000)
- 3Choose number of chains (default: 4)
- 4Select prior distribution (vague priors recommended for most analyses)
3
Run Bayesian NMA
- 1Click "Run Analysis" — the Bayesian sampling engine performs MCMC sampling
- 2Monitor convergence via trace plots and Gelman-Rubin diagnostics
- 3Review posterior distributions for each treatment comparison
- 4DIC helps compare model fit
Tip: If chains have not converged, increase burn-in or iterations.
4
Interpret results
- 1Probability of being best treatment for each intervention
- 2Rankograms show the full ranking distribution
- 3Posterior mean estimates with 95% credible intervals