Request Iterations to Converge PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Jaime Maerten-Rivera posted on Tuesday, October 06, 2009 - 11:27 am
I am wondering if there is a way to request the number of iterations that it took for a model to converge. Does anyone know of a way to do this?
 Linda K. Muthen posted on Tuesday, October 06, 2009 - 11:36 am
TECH5, TECH6, and TECH8 of the OUTPUT command provide optimization histories for different analyses. See the user's guide for further information.
 Jaime Maerten-Rivera posted on Friday, October 09, 2009 - 10:21 am
I used the TECH 5 command and I receive two sets of iterations. The first is the EM algorithm iterations for the H1 or unrestricted model. The second is the technical output from quasi-newton iterations with iterations using gradient, and iterations using quasi-newton.

Which model are the second set referring to and what is the difference between the two? I am trying to compare the number of iterations from HLM and SEM for a particular model, and don't know what to use.
 Bengt O. Muthen posted on Friday, October 09, 2009 - 5:07 pm
I assume you are considering an SEM. The EM algorithm is used with missing data to estimate H1. The gradient and QN iteration information that you see is for the H0 model. The HLM program does not compute H1 because it does not give a chi-square test of overall model fit.

Comparing the number of iterations across programs is a tricky thing to do. Depending on the algorithm there may be more/less iterations, but less/more time spent on each iteration. And the strictness of convergence criteria also plays in.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: