Negative binomial growth model PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Averdijk posted on Thursday, February 17, 2011 - 4:38 am
Dear Drs. Muthen,

I was wondering whether it is possible to estimate a negative binomial growth curve model in the current version of MPlus? I did not come across it in Chapter 6 of the User's Guide.

Many thanks in advance.
 Linda K. Muthen posted on Thursday, February 17, 2011 - 6:13 am
A count variable can be used in a growth model. Mplus can estimate a negative binomial model for count outcomes. See the COUNT option in the user's guide.
 Daniel Seddig posted on Thursday, June 20, 2013 - 5:08 am
Dear all,

i am working on a zero inflated negbin LGM for a count variable over 4 timepoints. A linear growth specification results in a negative mean slope, which fits the observed mean development of the counts. However, the Model Estimated Means reveal a totally different development (from t1 to t4: 2.491, 2.229, 4.093 & 14.779). Residuals are accordingly high. I am unsure how interpret this result. It could be a model misspecification, although nothing seems to point to this. Any thoughts on this? Thanks in advance.
 Linda K. Muthen posted on Thursday, June 20, 2013 - 7:56 am
Those values are log rates. The PLOT command plots the means which are the log rates exponentiated.
 Daniel Seddig posted on Sunday, June 23, 2013 - 1:19 pm
Hello Linda,

exponentiating these means does not result in reasonable values. Subtracting the residuals from the estimated means results in the observed means. So the initial estimated means do not seem to be log rates.

I would prefer to send you the data and output file for support. Thank you.
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: