Confirmatory LCA PreviousNext
Mplus Discussion > Categorical Data Modeling >
 Wen-Hsu Lin posted on Tuesday, January 28, 2014 - 1:04 am
I have 9 items want to do confirmatory LCA; however, I want to specify that some items should be related to one class but not another, based on theory. I have following syntax:
Names are
id g1 g2 g3 g4 g5 g6 g7 g8 g9;
IDvariable = id;
Usevariables are g1 g2 g3 g4 g5 g6 g7
g8 g9;
Categorical are g1 g2 g3 g4 g5 g6 g7
g8 g9;
classes = c(2)
Type = mixture;

[g1$1*1 g2$1*0 g3$1*1 g4$1*1 g5$1*0
g6$1*0 g7$1*1 g8$1*0 g9$1*0];
[g1$1*0 g2$1*1 g3$1*0 g4$1*0 g5$1*1
g6$1*1 g7$1*0 g8$1*1 g9$1*1];

Is the syntax correct?
 Linda K. Muthen posted on Tuesday, January 28, 2014 - 12:02 pm
I think you want @0 not *0 if you want to fix thresholds to zero.
 JW Beck posted on Tuesday, May 29, 2018 - 10:01 am
I have a question regarding confirmatory LCA: I set class-specific start values for each indicator according to the exploratory solution so that only the class sizes need to be estimated. However, I realised that some cases are differently assigned - dependent on the size of the data file: If I create a subsample and run again the confirmatory LCA, certain cases (I can track them based on their ID) are assigned to different classes. According to my logic, the assignment of cases to classes should, however, solely depend on the response pattern - independent of any N the data set has. Do you have any explanation for this observation? Thank you very much in advance.
 Bengt O. Muthen posted on Tuesday, May 29, 2018 - 5:33 pm
The class assignment depends on not only the response pattern but also the estimates of all parameters in the model (not only parameters for the indicators).
 JW Beck posted on Wednesday, May 30, 2018 - 12:22 am
Thank you for your swift reply. Do you mean that I have to fix all the other parameters as well? But wouldn't this imply that the class sizes are predetermined? Or is there a better way to assign a large number of new cases to an existing class solution? Thank you.
 Bengt O. Muthen posted on Wednesday, May 30, 2018 - 2:45 pm
If the new cases can be viewed as drawn from the same population as the sample used to estimate the model, it is correct to fix all parameters when classifying the new cases.
 JW Beck posted on Thursday, May 31, 2018 - 4:58 am
This answer helped me a lot! Thank you once more.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message