Latent growth modeling PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Thierno Diallo posted on Tuesday, February 05, 2008 - 12:29 pm
Hi Dr Muthen,

I am reading the article :
Muthen, B. (1998). Second-generation structural equation
modeling with a combination of categorical and continuous latent
variables: New opportunities for latent class/latent growth modeling.

and at the same time I am learning from the imput and the output from Mplus web site. I would like to ask you few questions.
From penn8,
1)How to get Table 10.3: Probality of dependence Given Class?
For example, how to calculate
P(Dep| c=1) =.20
2)What is the meaning of :
a)
Intercepts
C#1 -2.461 0.311 -7.906 0.000
C#2 -3.379 0.595 -5.681 0.000

b)
What is the meaning of Category > 1 in

LATENT CLASS ODDS RATIO RESULTS

Latent Class 1 Compared to Latent Class 2

DEP94
Category > 1 0.548 0.218 2.510 0.012

Thank you in advance for your help.
 Bengt O. Muthen posted on Wednesday, February 06, 2008 - 6:24 pm
This is part of our Short Course material for the mixture days. Very briefly,

1) Probability = 1/ (1 + exp(-Logit)) where Logit = - threshold.
Odds = Probability/(1 - Probability)
Odds Ratio = Odds (class k)/Odds (class K) for K = 4

2a) Those are the intercepts in the latent class multinomial regression on covariates ("c on x") - see chapter 13 of the UG,

2b) Categories are 1, 2, 3... for polytomous outcomes. So 1 and 2 for binary outcomes. Therefore, with a binary outcome >1 means category 2.
 Thierno Diallo posted on Monday, February 11, 2008 - 8:01 am
Hi Dr Muthen,
Thank you for your answer it was very helpful. Today my question is about Local maxima in mixture modeling. What can we do with a local maximum problem. I have already incresed the number of starts (from 50 to 500), but I still have this message:

WARNING: THE BEST LOGLIKELIHOOD VALUE WAS NOT REPLICATED. THE SOLUTION MAY NOT BE TRUSTWORTHY DUE TO LOCAL MAXIMA. INCREASE THE NUMBER OF RANDOM STARTS.

Thank you in advance.
 Linda K. Muthen posted on Monday, February 11, 2008 - 8:11 am
You may need more random starts, for example, 5000 500. If increasing the random starts does not solve the problem, 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: