MCMC判断标准：Gelman-Rubin检验与自相关时间分析

MCMC收敛判据：Gelman-Rubin检验和自相关时间

Gelman-Rubin 检验

The Gelman-Rubin statistic evaluates how comparable two sets of Markov chain Monte Carlo (MCMC) simulations are by comparing their mean and median values. This serves as an indicator of chain similarity, approaching unity as convergence occurs. A higher degree of similarity thus enhances confidence in their representation of the true posterior distribution. All parameters must achieve a Gelman-Rubin statistic below MAXGR (typically set at 1.01) to be deemed 'well-mixed' or 'converged'.

Tz检验

Tz代表独立抽样的数量。因相邻链接之间具有高度的相关性关系，在马尔可夫链蒙特卡罗