About Latent Class PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Qiao Hu posted on Monday, September 19, 2011 - 12:23 pm
I have 5 variables: V D A E C and centered the score of five variables, that is to say, the means of five variables is zero.
Each variable correlate with the other four variables.
I want to see the distribution of the five variables in the different latent group (class 2, class 3, class4 etc), then choose the most suitable class.

First, I choose two classes:

(1)I get such a warning, how to solve it?

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


(2) Why I cannot find Chi-Square Test of Model Fit in the output?

(3)The AIC BIC is negative, how to do with it?

Akaike (AIC) -15602.828
Bayesian (BIC) -15452.572
Sample-Size Adjusted BIC -15535.180
(n* = (n + 2) / 24)



This is my syntax.

DATA: FILE='C:\la.dat';
listwise= 0n;


VARIABLE: NAMES ARE V D A E C;
CLASSES = c(2);


ANALYSIS: TYPE = MIXTURE;


MODEL:
%overall%
V WITH D A E C;
D WITH A E C;
A WITH E C;
E WITH C;



OUTPUT:
TECH8 SAMPSTAT RESIDUAL STANDARDIZED;
 Linda K. Muthen posted on Monday, September 19, 2011 - 5:16 pm
1. Increase the random starts, for example,
STARTS= 200 50;

2. Chi-square and related fit statistics are not available when means, variances, and covariances are not sufficient statistics for model estimation.

3. This is not a problem.
 Qiao Hu posted on Monday, September 19, 2011 - 5:37 pm
Dear Linda,

This time I get another warning when I use START 200 50;

How to do with it?

WARNING: THE BEST LOGLIKELIHOOD VALUE WAS NOT REPLICATED. THE
SOLUTION MAY NOT BE TRUSTWORTHY DUE TO LOCAL MAXIMA. INCREASE THE
NUMBER OF RANDOM STARTS.
 Linda K. Muthen posted on Monday, September 19, 2011 - 6:37 pm
Increase the starts to 500 125. You need to increase the starts until the best loglikelihood value is replicated.
 Qiao Hu posted on Tuesday, September 20, 2011 - 3:20 pm
Dear Linda,

This time I increase the number of variables from 5 to 7. I get such a warning, how to do with it?

THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE
TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE
FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING
VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE
CONDITION NUMBER IS -0.531D-21. PROBLEM INVOLVING PARAMETER 43.


Best,

Qiao
 Qiao Hu posted on Tuesday, September 20, 2011 - 3:33 pm
Another question:

How big STARTS I should use? Because I increase the starts to 2000 1000, it still cannot work.
 Linda K. Muthen posted on Tuesday, September 20, 2011 - 4:13 pm
Please send the relevant files 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: