Posterior (likelihood × prior, conditioned on observed data)
Proposal Distribution
Ready
Press Run or Step to begin sampling.
Trace: θ&sub1;S = 0
Trace: θ&sub2;S = 0
Autocorrelation Function (chain 1, post warm-up)
Gelman-Rubin Convergence
DiagnosticsIteration: 0
Accept Rate
0%
target: 23–44%
Effective Sample Size
0
target: >400
R̂ (split)
—
target: <1.01
Diagnostic Guide
✓ Good mixing
All chains overlap in traces. Autocorrelation decays within ~20 lags. R̂ < 1.01. Effective sample size > 400. Accept rate 23–44%.
⚠ Slow mixing (σ too small)
Accept > 90%. Chain takes tiny steps — traces drift. Autocorrelation stays high for many lags. Effective sample size much less than S. Fix: increase σ.
✗ Stuck (σ too large)
Accept < 5%. Chain frozen with rare jumps — traces flat. Autocorrelation near 1 at all lags. Fix: decrease σ.
✗ Trapped (bimodal)
Chains stay in different modes. R̂ >> 1 because between-chain variance B dominates within-chain variance W. Fix: reparameterise or tempering.