How do you diagnose MCMC convergence?
hardAnswer
- (1) R̂ (Gelman-Rubin) < 1.01 across multiple chains → chains agree.
- (2) ESS (effective sample size) ≥ 400 per parameter.
- (3) Trace plots: should look like fuzzy caterpillars, no drift or stickiness.
- (4) Autocorrelation plots decay quickly.
- (5) Divergences in NUTS: indicate hard posterior geometry — reparameterize (non-centered) or increase .
- Never trust a single chain's samples without these checks.
Check yourself — multiple choice
- Never check
- R̂ < 1.01, ESS ≥ 400, trace plots caterpillar, quick autocorr decay, zero NUTS divergences; else reparameterize / longer runs
- Only ESS
- Random
MCMC diagnostics: R̂ / ESS / traces / autocorr / divergences.
#bayesian#mcmc
Practise Statistics Fundamentals
215 interview questions in this topic.