Multiple group multiple cohort growth... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Nick Mumford posted on Thursday, August 30, 2012 - 4:41 pm
Hello Mplus community.

My colleagues and I have a query regarding growth modelling.

In brief, our project has been to assess motor coordination of children aged 6-10yrs at 4 time-points over 2 years.Our aim is to model the development of motor coordination in children over time.

We have identified the ‘multiple group multiple cohort growth model’ described in section 6.18 of the Mplus user’s manual as the one that best fits the analyses we wish to perform.

However, while we can divide our participants up by age (as per the manual’s example) our age data is actually in months, rather than years. Thus, we have a more detailed set of age data to work with.

Our query is whether it is possible to do this kind of model using a more continuous variable like age in months, rather than clustering participants based on age in years.

Any advice would be greatly appreciated.
 Bengt O. Muthen posted on Friday, August 31, 2012 - 2:51 pm
You can certainly do a more fine-grained time division - that just leads to more cohorts/groups.
 Nick Mumford posted on Tuesday, September 04, 2012 - 4:29 pm
Thank you for your response Bengt. I think as a first step I'll attempt a larger time division, and then progress from there.

Just as a more general follow-up, as a new Mplus user I was wondering what the best avenue was for getting assistance with using the program and running models.

As I mentioned, I’ve isolated the model that best fits our aims and data, but making sure the next step of actually running the analyses is done properly may require some further assistance.

Thanks again.
 Linda K. Muthen posted on Wednesday, September 05, 2012 - 8:58 am
You have product support available at support@statmodel.com. Mplus Discussion is good for short questions that fit in one window. Neither Mplus support nor Mplus Discussion can provide statistical consulting. If you need this type of help, you should go to a consultant on your campus or post your question on SEMNET or a general discussion forum.
 Nick Mumford posted on Wednesday, September 05, 2012 - 5:49 pm
Thanks for that info Linda. I'm currently working my way through the lecture videos you've posted on the Mplus site, which have been very helpful in getting started.

That said, as a psychologist (rather than a statistician), I will likely have further questions as things progress, so it's good to have some options to follow up when the time comes.
 Nick Mumford posted on Tuesday, September 11, 2012 - 8:38 pm
Hi Linda and Bengt,

Just wanted to follow up with a question regarding the model we've discussed above.

On running the model I've received an error message mixed in with the results as follows:

WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IN GROUP C12 IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/ RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE S.

This error is repeated 12 times in the output, once for each of our 12 groups.

I'm struggling to know what the next step to take here is, and whether this means that the model results we've gotten should be discarded.

I've seen this thread (http://www.statmodel.com/discussion/messages/9/6843.html?1331645477) touched on this issue, but any advice would be greatly appreciated.
 Linda K. Muthen posted on Wednesday, September 12, 2012 - 6:56 am
Please send the output and your license number to support@statmodel.com.
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: