Message/Author 

Dexin Shi posted on Saturday, February 02, 2013  6:33 pm



Hi Dr. Muthens, Is there any way I can get Bayesian Credibility interval at a userdefined level (instead of 95% by default)? Thanks for your help. 


Not automatically. But you have access to the data behind the posterior distribution that you see in the plot and you could then find the intervals you want. 

Dexin Shi posted on Friday, March 01, 2013  2:29 pm



Thanks, Dr. Muthens; (1) is there any way I can save/get access the chains of values for the posterior distributions of the parameters? (I tried to look at the SAVEDATA command, but cannot get it work); (2) I was also wondering can I get access/save the bootstrap samples to an external file using Mplus 7. Thanks for your time. 


1. See the BPARAMETERS option in the Version 7 user's guide. 2. No. 

Dexin Shi posted on Friday, March 01, 2013  8:11 pm



Then using bootstrapping, can I get userdefined confidence interval levels other than the level given by CINTERVAL (BCBOOTSTRAP) (i.e.90%, 95%,99%)? Thanks. 


BCBOOTSTRAP is not available with Bayes. See pages 727728 of the Version 7 User's Guide for a description of the intervals available with Bayes. 


In the output from ESTIMATOR = Bayes; you get 95% credibility intervals given next to the point estimate by default. You can also get 90, 95, 99% intervals in a separate table below by requesting OUTPUT: CINT; I know that the default is for equal tailed intervals. But if you specify CINT(HPD); does that change the 95% credibility intervals given next to the point estimate to HPDs as well as those given in the separate table? Or are the 95% credibility intervals given next to the point estimate always equaltailed? 


Requesting CINT(HPD) changes the results in both places. 


I'm interested in estimating the width of the 95% credibility interval in a Bayesian SEM aggregated over MonteCarlo samples. Would taking the difference between the MonteCarlo point estimates of the 2.5th and 97.5th percentiles do this? Or is it more subtle than that? 


It is more subtle. We don't have an easy way to do this within Mplus. You would need either the Mplus automation or some other batch environment to do this in a Montecarlo but you have to save the confidence interval from each run and then average those. 


That's what I needed, thanks. 

Back to top 