Message/Author 


I will run weighted LCA analyses using NSDUH data. How do I take into account the complex sample NSDUH dasign in MPlus? There's a weight variable, a strata variable and a psu (primary sampling unit)variable. I believe that I can add the weight variable just by specifying the Weight in the input file. But how do I add strata and psu? Do I need or not to use Type=complex? Thanks, Silvia Martins 


There is a discussion of the complex survey data features of Mplus in Chapter 14 of the user's guide. I think this will answer your questions. You do not need TYPE=COMPLEX if you have only a weight variable. You need either TYPE=COMPLEX or TYPE=TWOLEVEL if you are using other complex survey data features. 


Thank you Linda! I read the discussion and it answered my questions. Silvia 


Linda After reading Chapter 14, I decided to ask my research assistant to ran a model using NSDUH data including weight, strata and psu. She's getting errors using the statification and cluster options. When she types both into the Variable command (as below), she gets an error, "unknown option: stratification". Additionally, when she takes out the statification option she gets the error "The labels %WITHIN% or %BETWEEN% must be specified before class labels". How does she solve this? Is it really necessary to add strata and psu when we just want to have sample weighted estimates? Thanks, Silvia 


Linda As an extension of the former message, this is one of the models than she ran (last message was two long so I split it in 2): IINPUT INSTRUCTIONS Title: Alcohol NSDUH LCA; Data:File is F:\GSKalcohol\nsduhalcohol2.txt; Variable: NAMES ARE realid weight vestr verep getuse kplim toleranc cutdown ctprb lessact withd; USEVARIABLES ARE getuse kplim toleranc cutdown ctprb lessact withd; Missing are all (9999) ; Classes= c(1); Categorical= getuse kplim toleranc cutdown ctprb lessact withd; Weight= weight; Within are getuse kplim toleranc cutdown ctprb lessact withd; stratification is vestr; cluster= verep; Analysis: TYPE = complex twolevel mixture; MODEL: %overall% %c#1% [getuse$1*1.4 kplim$1*1.4 toleranc$1*1.4 cutdown$1*1.4 ctprb$1*1.4 lessact$1*1.4 withd$1*1.4]; Savedata:File is F:\GSKalcohol\weightps1.txt; Save=cprob; Plot: Type is plot3; series is getuse(1) kplim(2) toleranc(3) cutdown(4) ctprb(5) lessact(6) withd(7); *** ERROR in Variable command Unknown option: stratification 


We ask that Mplus Discussion posts not exceed the space available in one window and not include input and output. Please send the input, data, output, and license number to support@statmodel.com and I will be happy to look at it. 


Linda Is there a maximum number of observations that MPlus can handle? For instance, if I have a datset with 110,000 individuals and 7 variables will Mplus be able to run LCA analyses? Thanks, Silvia 


There is not maximum to the number of observations in Mplus. 


Hi, Can I plot weighted longitudinal survey data? Does Mplus automatically produce a weighted plot when weight = is specified? Thanks. 


If the WEIGHT option is used, the weighted results are plotted. 

Cecily Na posted on Friday, February 15, 2013  12:45 pm



Hello professors, I am working on a national survey with strata (no clusters). I plan to use only subpopulation defined by two variables (majority individuals + living in metropolitan areas). I also need to use weights. So I should use type=complex? However, my analysis is a twolevel model, with variables associated with individuals and stratumlevel variables. How can I take that into account? Also when subpopulation is defined by condition A and B, what's the syntax? (Subpopulation is condition 1= A and condition 2 =B; ?) Thank you very much for your time! 


Perhaps you want to do a multiplegroup analysis using type=complex (for the weights), where the groups are defined by the strata. 

Cecily Na posted on Friday, February 15, 2013  4:23 pm



Thank you Bengt. A followup question. When I am only interested in a subpopulation, can I just use the subsample with unchanged weights? Thank you. 


Yes. 

Back to top 