I've performed a latent class analysis with R3STEP regression on multiple variables from a survey with a complex sampling design. I'm using CLUSTER and SUBPOPULATION to cluster at the classroom level and select out males or females for separate analyses. Some, but not all of the stratification variables from the survey design are included as covariates.
It seems from the User's Guide that I cannot also use weights. Is this the case? If so, what would be a good way to talk about the limitations of using the auxiliary variable approach with CLUSTER and SUBPOPULATION? What other approaches might be good? My goal is to evaluate the contribution of multiple chosen predictors (4 continuous covariates and a 4-level categorical variable) and significant interactions while controlling for two 4-level categorical demographic variables and two binary demographic variables.
Thank you for that input. I guess I don't understand what is meant by "The WEIGHT option is used to identify the variable that contains sample weight information for non-clustered data and level 1 for TYPE=TWOLEVEL and TYPE=THREELEVEL." on page 558 of the User's Guide. Since I am using clustered data and am not using TWOLEVEL or THREELEVEL, but TYPE=COMPLEX MIXTURE, what are the implications for that given the User's Guide text?