MCMC Diagnostics and Visualization

When fitting a model using the MCMC method (via emcee), the application provides a set of dedicated plots and diagnostics tools to analyze the posterior distributions of the parameters.

1. MCMC Fit Plot

  • Displays the median fit curve obtained from the posterior samples.
  • A 99% confidence band (credible interval) is shown around the curve.

2. Corner Plot

  • Shows the marginal and joint distributions of all model parameters.
  • Each subplot provides histograms (diagonal) and 2D scatter plots (off-diagonal).
  • Useful to assess parameter correlations, asymmetry.

corner plot

3. Walkers Plot

  • Plots the value of each parameter across sampling steps for all walkers.
  • Helps diagnose convergence and detect sampling issues (e.g., stuck walkers).
  • Walkers should appear well-mixed and stable after the burn-in period.

walkers plot

4. Autocorrelation Plot

  • Displays the autocorrelation time of each parameter.
  • Allows evaluation of sampling efficiency and helps determine appropriate thin or burn settings.

autocorrelation plot

Notes

  • All MCMC diagnostics are computed after discarding the burn-in period.
  • If convergence is poor, consider increasing the number of walkers or steps.
  • You can export posterior samples or corner plots from the toolbar menu.