Effective weights in Growth Mixture M... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 jennifer hamilton posted on Monday, November 03, 2008 - 3:55 pm
I would like to use effective weights (that sum to the effective sample size) in my growth mixture model. However, the only weighting option seems to be WEIGHT= (which rescales my weights to relative weights). I also thought about using the CLUSTER option, but this doesn't work for TYPE=MIXTURE.

Is there a way to incorporate effective weights into a GMM?
 Tihomir Asparouhov posted on Monday, November 03, 2008 - 4:17 pm
The MLR estimator does not depend on the scale of the weights, i.e., you will get the same results with effective weights or any other rescaled weights.

To use the CLUSTER option you need to specify in the input TYPE=MIXTUE COMPLEX.
 jennifer hamilton posted on Tuesday, November 04, 2008 - 12:11 pm
Tihomir,

Thank you for the prompt reply. If I could ask two follow up questions...

(1) I am not clear on why the MLR estimator doesn't care about the scale of the weights. Can you briefly explain, or point me in the right direction?
(2) What does the cluster option do in terms of the model equation? Does it include the cluster variable as a covariate in the model?

Thanks again for your continued patience,
Jen
 Tihomir Asparouhov posted on Tuesday, November 04, 2008 - 1:31 pm
Detailed information on the pseudo maximum likelihood estimator (MLR) is available in
http://statmodel.com/download/webnotes/mplusnote72.pdf
Formula (2) in that article shows how the scale of the weights cancels out. The cluster option accounts of similarities in the observations obtained from the same cluster. Generally speaking it adjusts the standard errors upwards.
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: