Mixture growth modeling on sequential... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Shige Song posted on Monday, March 20, 2006 - 12:32 am
Dear Linda and Bengt,

Are there good references on mixture modeling on sequential processes and how to implement it using Mplus?

For example, I have a goup of children. Each children's height was measured monthly during the first year after birth; then they were followed up at age 6, 7, and 8. I want to know:

1) the relationship between growth trajectory in the first year and that in the following years;
2) whether this relationship vary among some observed groups (sex, urban/rural, racial group, etc.);
3) whether this relationship vary among unobserved groups.

It appears to me that there are more than one possibilities. I can:

1) model the two processes separately (two sets of intercepts and slopes), and specify one latent categorical variable for the whole process (the first year and the following years);
2) or I can model the two processes separately and specify a latent categorical variables for each process, and model the transition probablity between the latent statuses (similar to example 8.7 in Mplus 4 online manual).

Question is: which one is better, and why? Also, are there other possibilities?

Thank you very much!

Best,
Shige
 Bengt O. Muthen posted on Monday, March 20, 2006 - 7:59 am
3 alternatives come to mind.

a. Do a regular (single-class) model where the growth factors for the 2 processes correlate (or where the growth factors of the later process is predicted by those of the former). This is preferred when it is not meaningful to work with classes. The ref. under b. below, however, points out that this alternative can be difficult to work with.

b. Summarize the development by a latent class variable as in Muthén, B., Khoo, S.T., Francis, D. & Kim Boscardin, C. (2003). Analysis of reading skills development from Kindergarten through first grade: An application of growth mixture modeling to sequential processes. In S.R. Reise & N. Duan (eds), Multilevel Modeling: Methodological Advances, Issues, and Applications (pp. 71-89). Mahaw, NJ: Lawrence Erlbaum Associates (#77). This is your alternative 1).

c. Use 2 latent class variable, one for each process, as in your alternative 2). If you send me an email, I will send you a submitted paper using this alternative.

I find that when it is meaningful to work with latent trajectory classes, alternative c. is quite useful because of its flexibility.
 Shige Song posted on Monday, March 20, 2006 - 8:27 am
Dear Bengt,

Thanks for pointing out some very helpful referencs, I will take a good look at it (I happen to have that book). I just sent you a message asking for paper that used alternative c.

Best,
Shige
 Daniel Rodriguez posted on Tuesday, November 28, 2006 - 5:07 am
Hi Linda and Bengt,
We received a peer review that highlighted the difficulty of reporting results from a sequential process gmm. Do you have any references or ideas on how to report the results of a sequential process gmm?
 Linda K. Muthen posted on Tuesday, November 28, 2006 - 12:27 pm
I am looking into a reference for you.
 Daniel Rodriguez posted on Wednesday, November 29, 2006 - 8:54 am
Thanks
DR
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: