Guidance/advice for a growth mixture ... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 LR posted on Friday, March 01, 2013 - 9:42 am
Hi !

I would like to ask for some advice/guidance. I have been using generalised linear mixed models for a particular dataset without much success. After some reading I think that a growth mixture model might be appropriate and that Mplus is the software I should use (which I have just acquired a license for).

The data come from a study of a substance abuse cessation attempts. 100 participants report their use on 10 occasions including baseline, after which the cessation attempt begins.

Clinical interest lies in two areas. First, is there a divergence in subsequent use (that is, lapse/relapse) between participants who are diagnosed with a particular disorder at baseline and those who are not ? Second, are there any distinct usage patterns/trajectories (classes) evident within the study population, and is membership of these classes predicted by the diagnosis aforementioned ?

The outcome variable is zero inflated since many of the participants do not lapse/relapse.

Potential confounders, both time varying and time invariant are also measured.

There is some loss to follow-up and I would like to handle the missing data appropriately (in the glmm framework I was using multiple imputation).

Is Mplus suited to this kind of analysis ? If so, I would be grateful for some guidance and/or links to any online examples of similar models.

Thanks
LR
 Linda K. Muthen posted on Friday, March 01, 2013 - 6:05 pm
It would seem Growth Mixture Modeling could be used for these data. See the criminology example in the following paper which is available on the website:

Muthén, B. & Asparouhov, T. (2009). Growth mixture modeling: Analysis with non-Gaussian random effects. In Fitzmaurice, G., Davidian, M., Verbeke, G. & Molenberghs, G. (eds.), Longitudinal Data Analysis, pp. 143-165. Boca Raton: Chapman & Hall/CRC Press.

See also the GMM papers on the website under Papers and the examples in Chapter 8 of the user's guide.
 jeon small posted on Friday, March 01, 2013 - 6:53 pm
The journal editor requested that I perform a post-hoc power analysis. The sample size is 438. The results from the measurement model are: X2=95.85, DF=40, probability level= .000, CFI=.954, RMSEA=.050

What do I need to know?
 LR posted on Saturday, March 02, 2013 - 5:38 am
Thank you Linda. In the paper you referenced it says that "input scripts for the analyses are available at http://www.statmodel.com" but I have not been able to locate this on the website. Please advise.
 Bengt O. Muthen posted on Saturday, March 02, 2013 - 4:30 pm
Turns out that I didn't post them. Mostly because they have counterparts in the User's Guide. Let us know if there is a particular model type that you don't find in the UG and I can dig into my runs.
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: