Latent variable as predictor of laten... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Nelli Hankonen posted on Wednesday, October 29, 2008 - 7:40 am
Hi,

I would be interested in three questions regarding predicting latent class membership. It would be great if someone could help!

In the Mplus manual of Version 5, in the example 8.1, the predictor x is a manifest variable. Could that predictor be latent?

Why are the intercept and variance of that predictor not estimated?

Also, in this example, the x is predicting not only the class, but also the intercept and the slope. Is it possible to predict the class membership only?

Thank you very much in advance!
 Bengt O. Muthen posted on Wednesday, October 29, 2008 - 8:04 am
Answering your questions in turn:

Yes.

Parameters related to predictors are not estimated just like they are not in regression.

Yes.
 Nelli Hankonen posted on Wednesday, October 29, 2008 - 8:21 am
Hallo,

many thanks for your answers.However, I am still a bit confused about 1. the estimated parameters of a latent class predictor and 2. the nature of latent class predictors:
1. In regression we know that although the means, variances, and covariances of predictors are usually not shown in an output they need to be estimated. Why did I have to set these to 0 when I was predicting latent classes?
2. I tried to predict class membership with a latent variable that had 5 indicators. Logics would have it that this factor has a mean and a variance (cf. 1.) and that its indicators have intercept 0 and residual variance. That did not work. Then I tried all combinations of setting the factor mean, the factor variance, the indicators´intercepts, and the indicators´ residual variances to 0 but nothing work. Is there an example of a latent variable predicting class membership (e.g., variable x in example 8.1 being latent)?

Again, thank you very much in advance!
 Linda K. Muthen posted on Wednesday, October 29, 2008 - 8:53 am
Please send your input, data, output, and 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: