Is it possible to achieve MCMC convergence diagnostics through iterative initialization?

Estimation ML/Bayesian estimation
4

Your MCMC shows poor mixing but not really a trend. I guess you will simply need a lot of draws to properly sample from the ergodic distribution.