Convergence diagnostic for Bayesian parameter estimation

This panel provides information about the quality of the obtained MCMC samples, by means of Gelman and Rubins (1992) scale reduction factor and traceplots for each parameter of interest. Note that a scale reduction factor close to 1 indicates good convergence. In addition, a good overlap between the chains in the traceplot indicates that the chains have converged from their initial starting values to their stationary distribution. If convergence cannot be assumed, try out a larger number of samples, a longer burn-in period or more thinning. Details about Bayesian cognitive modelling and MCMC methods can be found, e.g., in Lee and Wagenmakers (2013).