Message/Author 


Dear All, I am a new user of Mplus and currently using version 7. I am investigating the neighbourhood environment's influence people's activity patterns (walking behaviour). I have individuallevel (level1) and neighbourhoodlevel(level2) factors. I read a paper by Henry and Muthen (2010) "Multilevel Latent Class Analysis: An Application of Adolescent Smoking Typologies with Individual and Contextual Predictors" that used MLCM. I am considering using that approach in my data analysis for this reason: to identify the homogeneous neighbourhood typologies and as well as accounting for nested structured in the data set. Is it appropriate to use the latent classes in my analysis as the main exposure variables and walking behaviour as my outcome variable? Sincerely, Ernesto. 


The neighborhood typologies would be captured by cb, a betweenlevel latent class variable. Using cb as exposure and walking behavior as outcome is accomplished by the mean of the betweenlevel part of the walking behavior variable varying across the cb classes. 


Hello Dr. Muthen, Thank you for your quick response and explanation! I would like to know if there is any literature other than Henry and Muthen (2010) or excerpt of a book chapter available online to familiarise myself with? That is, where the author used for example "cb" as an exposure variable and individuallevel factor (for example, activity patterns) as an outcome variable? Sincerely, Ernesto. 


No. But the UG has examples where a latent class variable influences a distal outcome. See e.g. UG ex 8.6. You can piece it together starting from that. 


Dear Dr Muthen, Thank you for the suggestion. I just identified that the example 10.2 in the UG "Twolevel mixture regression for a continuous dependent variable with a between categorical latent variable"? may be able to assist my work. My reason is that I am investigating on how neighbourhood environmental features influence a person's activity patterns, which means that I should be focusing on latent classes at the neighbourhood level instead of individuallevel. In that example, both levelone and leveltwo factors are all continuous variables. What case in point may fit a situation where the dataset has both categorical and continuous variables? Thank you for your support! Sincerely, Ernest. 


That combination presents no special input difficulties. Just put the categorical variables on the Categorical list. 


Thank you, Dr. Muthen for your support. 


ok. 


Hello Dr. Muthen, I am practising how to get the following plots; i.e., either figure 8 or figure 9 in Henry and Muthen (2010). Is there a specific Mplus code to get the bar chart in either figure 8 or figure 9 having both levels on the same plot? Thank you for your support! Sincerely, Ernest. 


No, this was done outside Mplus. 


Okay! Thank you, Dr. Muthen. 


Hello sir, I am using the 7.4 demo version and I am doing a multilevel lca. I have a question to ask sir. I have 6 indicators, student level covariate which is gender, and courses as my level 2 or cluster. ANALYSIS: TYPE = TWOLEVEL MIXTURE; STARTS = 100 50; MODEL: %WITHIN% %OVERALL% %BETWEEN% %OVERALL% C#1; is my model correct sir? what does c#1 mean? 


See Example 10.4 in the user's guide. 


I'm sorry sir but can you explain it to me. I am new to Multilevel lca and this is my first time using MPlus. As from what I've understand describes the student and course level. But I have no idea what is C#1. %WITHIN% %OVERALL% %BETWEEN% %OVERALL% Am I doing the right analysis sir? 


If you are new to mixture modeling, you should the Topic 6 course video on the website where all of this is described. 

Djangou C posted on Monday, May 22, 2017  4:23 pm



Dear Mplus team, I am running multilevel latent class analysis and I have some questions regarding which model to use. According to Henry and Muthen (2010), indicators specific random effects can be included in the formation of the latent classes. I was wondering if there was any rule regarding their inclusion in the model. Do we know the consequences of omitting these indicators in class formation and parameter estimates? I couldn’t locate any paper that discussed this issue. Thank you for your help. 


No rules that I know of. But if these indicatorspecific variances are significant I would include them. 

Back to top 