TECH14 with mixture complex PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Bernice Garnett posted on Thursday, July 19, 2012 - 3:18 pm
Hi,

I am getting a warning message that Tech 14 is not available for Type = complex. I am running an LCA with complex survey data (kids nested in schools) and so I have been using Type = mixture complex to account for the non-independence of observations within schools.

I wanted to compute a parametric bootstrapped test because I have sparse data.

Do you know how I can do this while also accounting for clustering with a type = mixture framework?

Many Thanks
 Linda K. Muthen posted on Thursday, July 19, 2012 - 4:57 pm
Please send your output and license number to support@statmodel.com.
 J.D. Haltigan posted on Wednesday, August 06, 2014 - 2:51 am
Hello:

I am also getting this same warning and was wondering if there was a known reason for it? My syntax is virtually identical to that in Appendix C of the Nylund dissertation as I am also working with peer victimization data.

Many thanks
 Linda K. Muthen posted on Wednesday, August 06, 2014 - 3:36 pm
TECH14 has not been developed for complex survey data.
 J.D. Haltigan posted on Wednesday, August 06, 2014 - 5:51 pm
To be sure in the Nylund et al., 2007 CD paper, she mentions using a sandwich estimator to account for the nonindependence of students nested within schools. Cross-referencing this with her dissertation she notes in a footnote that the (nonindependence) adjustment is specified using 'type = complex'. Since the 2007 CD paper reports the BLRTS for different class models and Appendix C in the dissertation shows TECH14 along with Type=mixture complex, it is not clear to me in the 2007 CD paper how nesting was handled (since BLRTS are reported which can not be obtained in TYPE=MIXTURE COMPLEX) in the different class runs for peer vic. Am I misreading it or was nesting handled in a different way in the 2007 CD paper?
 Bengt O. Muthen posted on Wednesday, August 06, 2014 - 8:31 pm
Type = Complex affects the estimates when weights are included and affects the SEs due to weights, stratification and clustering. The Tech14 BLRT is not developed to take these features into account due to the complexities of doing the bootstrapping with weights and clustering.
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: