Message/Author 


I understand the default in Mplus is to discard half of the iterations as burnin. Could I specify something different, a specific number of iterations or a different ratio (e.g., 1/10)? I am working on a model that converges very fast, but some of the parameters have large autocorrelation, so I would like to use a big thinning factor and run long chains. This means the burnin I need is only a small fraction of the number of actual iterations. Thanks! 


Auto correlations are not a reason of concern in principle. I would recommend to not alter Mplus convergence decision and posterior distribution. If the auto correlations is extremely high the MCMC sequence can not possibly converge fast. I would recommend using thin as you have done but leave the rest as is. You can save all parameters in all MCMC iterations with the BPARAMETERS and summarize the posteriors any way you want from that file. 


Thank you, Tihomir. 


Hi Tihomir, I just wanted to share a bit more. I used four chains and set BCONVERGENCE = 0.001, and convergence happens within a few hundred iterations. This is very clear from the trace plots, where the four chains quickly converge and mixing is very good. Most of the parameters have low autocorrelation, except for binary variables' thresholds and regression coefficients relating these to a latent mediator. Autocorrelation is not a concern except that it reduces effective sample size, and I was trying to get to effective sample sizes that I feel comfortable with (a few thousand). Yes, I have saved all the iterations and discarded only part of the first half. I was just wondering whether there was a way to specify a burnin option so you don't have to do this manually. Thank you. 


PS: Sorry, I misspoke. In that model I have an observed continuous mediator, which is why it converges very fast. I have another model with a latent mediator underlying an ordinal variable; that one takes a longer time to converge. 


Trang Use FBITER command to specify whatever number of iterations you want. BCONVERGENCE = 0.001 is a very strict criterion and I have no doubt the model has converged. Models with categorical variables in general tend to take longer to converge and have higher auto correlations. Tihomir 


If the BCONVERGENCE criterion is set to .001, what is the PSR boundary for convergence? I know that the PSR boundary is based on the number of parameters estimated within the model. Within this note: http://www.statmodel.com/download/Bayes2.pdf on page 8, some text indicates that convergence is reached when PSR values for all parameters is less than 1 + e, where e = f*c. c is set by the user (via BCONVERGENCE = c), but what is f? How can we determine f? Thanks! 


You should use FBITER if the automatic setup is not working for your models. It should work for most standard models. In excel f=NORMINV(0.95^(1/p),0,1)/1.64485362695147 where p is the number of model parameters. Thus if p=1 then f=1. 

Daniel Lee posted on Thursday, January 10, 2019  10:08 am



Hi Tihomir, Can you very briefly explain why categorical variables have higher autocorrelations than conntinuous variables? Thank you so much! 


I assume that your question is why you get smaller correlation when a categorical variable is treated as continuous. This is wellknown in the literature and it is generally referred to as attenuation. It happens not just in time series models but also in crosssectional models. It has been documented first here http://www.statmodel.com/bmuthen/articles/Article_036.pdf It is also discussed in this recent article http://www.statmodel.com/download/CenteredMediation.pdf under Table 9. 

Daniel Lee posted on Monday, January 14, 2019  9:33 am



Thank you! 

Back to top 