

LCA and longitudinal analysis 

Message/Author 

nina chien posted on Thursday, May 08, 2008  11:14 am



Is it possible to: 1) use one set of indicators (i.e., parenting) to form latent profiles, 2) then use these latent profiles to predict growth trajectories of a second set of indicators (i.e., child outcomes) (So I think this is neither latent class growth analysis nor growth mixture modeling) If so, can you refer me to any references? Thank you for you help. 


Yes. See the following paper which is available on the website: van Lier, P.A.C., Muthén, B., van der Sar, R.M. & Crijnen, A.A.M. (2004). Preventing disruptive behavior in elementary schoolchildren: Impact of a universal classroombased intervention. In Journal of Consulting and Clinical Psychology, 72, 467478. 


Dear Prof. Muthen, For a sample of size (i), I have five(t=5) binary response variables, Y(it), and each binary response variable is a function of a vector (four of them) of scenariodependent covariates X(it) and scenarioindependent covariates Z(i). In other words, Y(it)= f[X(it),Z(i)]. Additionally, Y’s are collected from a sequence of 5 question that were asked to respondent in survey, that is there is no time difference between Y’s except that they came in a sequence ( i.e. NO growth model)and therefore they may share some autocorrelation. I want to estimate a latent class (logit) model. Can I use Mplus to estimate this? Could you suggest me one example. I use version 4.1 Thanks and regards 


Yes, you can use Mplus to do such a LCA. You would have direct effects of x and z on the items  those effects could even differ across classes. And you could, if you need to, include a factor in the model to account for the "autocorrelation" you mention. The version 5 UG which you find on our web site has several examples. 

Back to top 

