Constrained factor loadings in Estima... PreviousNext
Mplus Discussion > Structural Equation Modeling >
 Phil Wood posted on Wednesday, July 06, 2011 - 2:45 am
Is it possible to specify a nonlinear constraint of the form:
in Mplus using Bayesian estimation?
(For example, assume that P1, P2, and P4 are all factor loadings.)
If not- do you recommend using phantom variables?
 Bengt O. Muthen posted on Wednesday, July 06, 2011 - 3:02 am
Not yet - only unconstrained NEW parameters are handled in Bayes so far. So, great if you can do it via phantom variables.
 Phil Wood posted on Wednesday, July 06, 2011 - 5:25 pm
Is there any way to get more decimals of precision for the posterior predictive P-value? I'm interested in this because I want to calculate Bayes factors for model comparisons.
 Tihomir Asparouhov posted on Wednesday, July 06, 2011 - 7:40 pm
You can use the fbiter command with a large value, for example fbiter=50000. That will yield more precision in the PPP value. Precision beyond the 3 decimal points will require huge number of iterations so it is not practical.
 Phil Wood posted on Thursday, July 07, 2011 - 1:52 am
Well, the problem is that all I have for the model I want to reject is 0.000 and the model I want to retain is 0.200, but I need the first few significant digits in the printout in order to calculate a bayes factor. (I know that the .2 is "good enough" to retain, but want to calculate the Bayes factor anyway.
Sorry to not have spelled it out in more detail the first time!
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message