Bayesian & MCMC Lab — Priors, Posteriors, Diagnostics | GetCalcMaster
Explore Bayesian inference with MCMC sampling. Inspect trace behavior, uncertainty, and diagnostics — then export posteriors and assumptions into the notebook.
Bayesian & MCMC Lab
The Bayesian & MCMC Lab is built for data-driven uncertainty: specify priors, combine with likelihoods, and sample posteriors using MCMC. The focus is on interpretation and diagnostics, not just point estimates.
What you can explore
- How priors influence posteriors.
- Posterior uncertainty summaries (credible intervals).
- Chain behavior: burn-in intuition, mixing, and convergence warnings.
Diagnostics mindset
MCMC can look “fine” while being wrong if chains do not mix or if you have too little effective sample size. Always inspect traces and run multiple chains when possible.
Keep assumptions and model choices in the notebook so the analysis is explainable and reproducible.
FAQ
What is the difference between a confidence interval and a credible interval?
Credible intervals summarize posterior probability under your model and priors. Confidence intervals are a frequentist construct based on repeated sampling properties. The interpretation differs.
How do I know if my MCMC converged?
Look for stable trace behavior, good mixing, and consistent summaries across chains. If estimates change substantially with longer runs, you likely need more sampling or a better sampler/model.
Can I use MCMC without much statistics background?
You can explore and learn, but for high-stakes inference you should validate assumptions and consult statistical references, especially around priors, identifiability, and model checking.