Hi, professor, my data with a complex sample of multistage cluster random sampling, i want to do a latent class model , and use first approach : compute standard errors and chi-square tests of model fit taking into account complex sampling features. This is my Language: DATA: file is "C:\Users\Administrator\Desktop\xwms.dat"; VARIABLE: NAMES ARE x1-x5 urdsp code fweight; USEVARIABLES =x1-x5; CATEGORICAL=x1-x5; STRATIFICATION=urdsp; CLUSTER=code; WEIGHT=fweight; CLASSES = C(3); ANALYSIS: type= mixture complex; OUTPUT: TECH1 tech11 ; TWO QUSETIONS: 1) Is this language all right? Is the CLUSTER means primary sample unit ? WEIGHT means final sample weight ? 2) what would you consider to be the "best" estimator for this model? Can pseudo maximum likelihood estimation method do (i read it in some literatures, but it can not be used in mplus ) ? or some method else ? Thank you very much!
CXN posted on Wednesday, November 01, 2017 - 8:01 pm
1)I tried to run my Language fore-mentioned, i find that even though delete the CLUSTER or STRATIFICATION from my Language,and just keep WEIGHT, the results showed nothing changed, does it means the CLUSTER or STRATIFICATION is insignificance for model, and why? 2)my data used a multistage sampling,should i take any statement about stage variable in this model ? 3)The "best" estimator for this model is ML or MLR ?
Looking forward to your reply and give my sincerely thanks to you!