Multiple group LCA with count& ordina... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Michelle Colder Carras posted on Friday, August 16, 2013 - 1:52 pm
Good morning Drs. Muthén and thank you for your excellent work.

I’m hoping you could help me understand how to take my understanding of Multiple Group Analysis for factor analysis to the same for a categorical LV modeled with an ordinal group indicator (age, 5 categories), 9 count variables (0-24) and 14 categorical (0-5) indicators.

So far I have thought to model with each member of the group (age) separately using the syntax below(for the first age category, age1), planning to constrain parameters for subsequent models (ages2-5). However, I wonder if the GROUPING option might be more appropriate? In that case I’m not sure what I would constrain for the count variables—means & variances? Would these be done in separate models, e.g. modeling age1, then age2, etc?? Also, which output would be best?

Thank you in advance for your help.

Variable:
Names are
yr school class ord1-ord15 educ count1-count9 age1;

Usevar are
ord1-ord15 count1-count9 age1;
Categorical are
ord1-ord15 age1;
Count are
count1-count9;
classes are pg(2);
Analysis:
TYPE = MIXTURE;

STARTS 100 20;
 Linda K. Muthen posted on Saturday, August 17, 2013 - 1:45 pm
With TYPE=MIXTURE, the GROUPING option is not available. You should use the KNOWNCLASS option which accomplishes the same thing. When all classes are known, it is the same as multiple group analysis.
 Michelle Colder Carras posted on Saturday, August 17, 2013 - 6:00 pm
Thank you for your response. I did try KNOWNCLASS and got the error message that since I have missing on the grouping variable, that is not an option. I saw from another post that you had recommended "run the model with the factors and variables shared by all groups and test invariance of the factors over groups for those factors. Then you would have to run the group separately that has more factors and variables", which I'm thinking means what I planned originally.

But as I try that I get a model that runs for 1.5 hrs with 2 messages about perturbed starting values not converging, and a message about "AN INSUFFICIENT NUMBER OF E STEPS. INCREASE THE NUMBER OF MITERATIONS." MPlus then added a MODEL statement using final estimates as starting values. Given that the three-class model for grouping category one won't even converge, what do you recommend next? Thank you again.
 Michelle Colder Carras posted on Saturday, August 17, 2013 - 6:02 pm
I'm so sorry; I should absolutely add that I omitted

Missing are all (-9999) ;


from my original post--but not the model syntax--to save space in my extra-long posting :-)
 Linda K. Muthen posted on Sunday, August 18, 2013 - 4:31 pm
Neither GROUPING or KNOWNCLASS can have missing values. You should sort that problem out and use the KNOWNCLASS option.
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: