Questions about LCA PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 aguri posted on Monday, November 19, 2012 - 10:04 pm
Dear Dr. Muthen
I am using MPLUS 3.11 to run a Latent Class Analysis.
The database are from TIMSS 2007 about student attitudes to math, 12 items(Likert 4-point) and about 4000 test takers included.I try to make exploratary LCA to find how many classes would be approriate, two questions are bumped into:
1. the p-value of Likelihood Ratio Chi-Square from c(1) to c(3) are 1, when c(4), something showed up...
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO A NON-POSITIVE DEFINITE FISHER INFORMATION MATRIX. CHANGE YOUR MODEL AND/OR STARTING VALUES.
.......

so I try to recode the data from 4-point to 2 categories (positive and negative orientation),and the result showed 5 classes is much better. I am not sure what happened when 4-point situations, and could I deal with the problem as description !!??(Do you recommend to change estimates method or setting starting value!?)
2. If I want to know which class level each test taker belongs to, what can I do !!??
THANKS A LOT :-)

my code as reference ~
TITLE: LCA
DATA: FILE IS test1115.dat;
VARIABLE: NAMES ARE gender i1-i12 ;
USEVARIABLES ARE i1-i12;
CLASSES = c (3);
CATEGORICAL = i1-i12;
ANALYSIS: TYPE = MIXTURE;
OUTPUT: TECH1 TECH7;
 Linda K. Muthen posted on Tuesday, November 20, 2012 - 12:00 pm
The two chi-square statistics given for the frequency tables of the categorical variables are not useful when you have more than about 8 variables. You should ignore these.

You should not recode your data. You should use more starts, for example, 1000 250.
 Sujith Ramachandran posted on Tuesday, November 15, 2016 - 10:56 pm
My LCA model returned a non-significant p value for the 4 class model, but the 3 class model seems to have identification problem. I increased the random starts to (1000 100), and the best loglikelhood replicated twice, but it displayed this error.


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 -0.181D-16. PROBLEM INVOLVING THE FOLLOWING PARAMETER:
Parameter 4, %C#1%: [ PERC_CASH ]


What does this error mean? And how do I fix it? Or does it suggest that I should move to the 2 class model instead?
 Linda K. Muthen posted on Wednesday, November 16, 2016 - 2:07 pm
Please send the output and your license number to support2statmodel.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: