Message/Author 


I'm trying to fit a logistic regression with latent classes and covariates that predict the latent classes. I've prepared the following syntax: TITLE: DIF analysis; DATA: FILE IS Data.txt; VARIABLE: NAMES ARE ID SES ethn PTS INT L1 CF FIELD FINC ITD ADMC DE VCBK GRMK READ NBR REST PRED EMR GRW RUBD MOT MCIN TME PEE OSEE ENRE ENFI ENNF LCT TER HOR ONR item1item24 gender Total GenTotal cprob1 cprob2 n; USEVARIABLES= SES item1item24 gender Total GenTotal n; nominal = n; MISSING = *; CLASSES = c (2); ANALYSIS: TYPE = MIXTURE; STARTS = 0; MODEL: %OVERALL% item1item24 ON Total gender GenTotal; c ON SES; %c#1% [N#1@3.261]; %c#2% [N#1@3.936]; item1item24 ON Total gender GenTotal; OUTPUT: TECH1 CINTERVAL; Is the syntax correct? If yes, which part of the output shows if the path from the covariate to the latent class is significant? 


Looks fine although I don't know why you have the fixed nominal parameters (3step?). C ON SES shows you the paths from SES to C. 


Thank you very much Dr. Muthen, It is in fact a 4step. The fourth step will have more than one covariate. Unfortunately, when I run the software, it given an error: THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN INSUFFICIENT NUMBER OF E STEPS. INCREASE THE NUMBER OF MITERATIONS. ESTIMATES CANNOT BE TRUSTED. I increased the miterations to 20000. Mplus gives the same error even with this number of iterations. The problem is I don't get the pvalues for the covariates. Is it because the model does not converge? What do you suggest? As for the fixed parameter, I put in the paramters from step two that come under this sections: Logits for the Classification Probabilities for the Most Likely Latent Class Membership (Row) by Latent Class (Column) Is that right? Many thanks for your kind help in advance. 


Check out our FAQ called TECH8... Yes, those are the logits you want to use. 


Following my previous questions, when we move from step 1 to steps 2 and 3, the latent classification of the individuals must remain fixed. That is, after the latent classes are identified in step 1 and individuals are assigned to a latent class, this cannot change when we add covariates in step 3. Right? How do we fix it in Mplus? (How do we tell the software that the latent classes and the number of participants in each class from step 1 is kept untouched?) I hope my question is clear. 


See the Mplus Web Note 15 and 21 on our website. 

Back to top 