View plots of the posterior parameter draws against their draw numbers or the log posterior kernel values versus their draw number. It is also possible to plot the normalized weights for the SMC posterior sampler with likelihood tempering and the importance sampler based on the MitISEM algorithm. You can choose between the original and the transformed parameters, while the log posterior kernel values are always computed for the transformed parameters. Below is an example that displays 100,000 posterior draws of the transformed parameters. The horizontal solid black line is the posterior mode estimate from the joint distribution, while the vertical solid black line is the end of the burn-in period. This occurs at 10,000 draws below. It can also be seen that the sampler (random-walk Metropolis with normal proposal density) displays convergence problems for the rA parameter at the selected sample size.
Additional Information
• | A more detailed description about convergence of the posterior sampler can also be found in Section 9 of the YADA Manual. |
Page url: http://www.texlips.net/yada/help/index.html?raw_posterior_draws4.htm