Message/Author 

Rob McGown posted on Tuesday, November 22, 2005  9:06 pm



I have some longitudinal Federal and longitudinal State data for early reading achievement. I'm seeking guidance on what is the most appropriate way to make valid comparisons between the growth curves of these data. What would be the best way to do this with Mplus? Can I just compare models estimated separately? Or would it be better to try to simultaneously estimate latent growth curves for the seperate outcomes? Thanks! 


Are the reading achievement measures for the data sets the same? 

Rob McGown posted on Wednesday, November 23, 2005  8:21 am



No....one is IRT scale scores and one is SAT9..they are different tests. Thanks! 


Then you should run them separately and it would not make sense to compare them in any absolute way. 

Rob McGown posted on Wednesday, November 23, 2005  9:26 am



can I compare growth rates? 

Rob McGown posted on Wednesday, November 23, 2005  9:31 am



what about standardized solutions, can they be compared? Thanks!!! 


I don't think any of these can be compared in an absolute sense. The same would go for standardized. 

george posted on Wednesday, March 25, 2009  5:58 am



Does it make sense to do a significant test for differences in intercepts and slopes among mixtures in GMM? If yes how is it done in Mplus? 


You can do this using the Wald test of MODEL TEST or loglikelihood difference testing. See the user's guide for MODEL TEST and Chapter 13 for loglikelihood difference testing. The difference test compares two models, one with equalities across classes and one without. 

Lorelai posted on Thursday, March 24, 2016  4:09 am



Hello, I would like to compare growth curves concerning 'study experience' (consisting of several variables) of two groups of students (1: students who drop out of their study or are delayed at some point during the first two years of their studies, 2: students who are still on track after two years). At four points in time during the first two years of their studies, we gathered 'study experience' data, but of course we 'lose' participants from the first group along the way. My first question is: do you agree that comparing the growth curves of these two groups is a sensible way to analyze these data taking into account the 'loss' of subjects in one of the groups? In addition, I am of course concerned about inequality concerning for instance student characteristics of these two groups. Therefore I deem it to be sensible to include some relevant covariates. Therefore, my second question is: how do I model the comparison of two growth curves with coviariates in Mplus? Many thanks in advance! 

Lorelai posted on Thursday, March 24, 2016  5:28 am



An addition: I am considering a two group latent growth model with time invariant coviariates (since they do not change over time) 


A twogroup growth model with covariates can be tried if the outcome is the same for the two groups. The big issue is the character of the dropout. Dropout in growth modeling is discussed in the paper on our website: Muthén, B., Asparouhov, T., Hunter, A. & Leuchter, A. (2011). Growth modeling with nonignorable dropout: Alternative analyses of the STAR*D antidepressant trial. Psychological Methods, 16, 1733. You may also want to pose this question on SEMNET. 

Back to top 