Message/Author 


For the following Multilevel CFA model: CATEGORICAL: answer1 answer2 answer3 answer4; ANALYSIS: TYPE IS TWOLEVEL; ESTIMATOR IS WLSMV; Model: %within% W_Factor BY answer1* answer2 answer3 answer4; FIXED VARIANCES W_Factor@1 %between% B_Factor BY answer1* answer2 answer3 B_Factor@1 You can see from the above model that I do not want to specify answer4 to load on B_Factor at the between level. My understanding is that variables have to be present at both levels, so how could the variable 'answer4' be handled? For continuious data using ML in EQS for example I would merely estimate the error variance for answer4 at the between level but because I am new to MPLUS and because treating data as categorical I need advice as to the appropriate way to proceed. 


Two alternatives. One is to allow answer4 to have betweenlevel variation, in which case you can simply say answer4; on the between level (this may even be the default) to estimate its betweenlevel variance. The other is to say that answer4 does not have betweenlevel variation which can be handled in two ways. One is to put answer4 on the within= list. The other is to say answer4@0; on the between level. 


Thank you Bengt. Changing the topic slightly, lets us say that for wahtever reason I wanted to remove the clusterlevel item specific erros for all 4 items then am I right in thinking that all I need to do to achieve this is to fix these to 0 at the between level? As per below: CATEGORICAL: answer1 answer2 answer3 answer4; ANALYSIS: TYPE IS TWOLEVEL; ESTIMATOR IS WLSMV; Model: %within% W_Factor BY answer1* answer2 answer3 answer4; !FIXED VARIANCES W_Factor@1 %between% B_Factor BY answer1* answer2 answer3 !FIXED VARIANCES B_Factor@1 answer1@0 answer2@0 answer3@0 answer4@0 


right. 

Back to top 