Message/Author 

Jan Ivanouw posted on Thursday, June 05, 2008  1:29 pm



Please, I am wondering about the following: What is the difference between Latent Class Analysis and the old method 'cluster analysis' (which for instance is implemented in SPSS)? 


LCA is a type of cluster analysis but not the same as what is in SPSS. See Chapter 3 of the following book: Hagenaars and McCutcheon (2002). Applied Latent Class Analysis. 


Thank you for the clarification Jan Ivanouw 


Dear Mrs. Muthen, I have a dyadic data set (N=157 couples) and I have two main goals: 1) create a profile, separately, for females and males based on 6 continuous variables. 2) see how these profiles predict several continuous outcomes. For 1) I tried: a) two step cluster analysis and latent class analysis. The results of the cluster analysis were more sound. So I have 2 clusters for females and males. To answer goal 2) i would like to insert the clusters as predictors of my independent variables. And finally test a mediation model inserting a continuous variable. I wonder how can I do this analysis? Is this a path analysis with categorical variables as predictors? If yes, where can I find an example of Mplus input. Thank you in advance. 


Using the latent classes in the modeling is like doing multiplegroup analysis. Eacb group/class can have its own parameters such as means, intercepts and regression slopes. Note that you don't say "y ON c". Instead, this implies that the y means vary across the c classes. So, the Mplus setup is straightforward. In addition to your latent class indicators you add the variables in your regression relationships and let those variable parameters vary across the classes. 

Back to top 