Hi, I am a new user of Mplus. We are studying changes over time in a set of related dichotomous symptom indicators (a given symptom is either present or not-present at each time point in a large sample of adolescents and adults).
Two questions are at the heart of our research. First, what is the latent trajectory of symptom status across the lifespan for each separate symptom? Second, which symptoms tend to change together?
I am beginning to have some luck fitting latent growth models with Mplus to address the first question.
However, I need help understanding whether Mplus can help us essentially do a factor analysis of latent symptom slopes. Any suggestions would be appreciated. I would also appreciate references to research that has addressed this kind of question.
You can consider doing an EFA followied by a CFA at each time point to determine whether you are measuring the same number of dimensions at each time points. You can then test measurement invariance over time. If you have measurement invariance, you can model the growth in the latent variables. I am not aware of references to this particular model.
I am new to Mplus and have a question. I have 3 time points for three continuous variables (say a1,a2,a3; b1,b2,b3; c1,c2,c3).I ran CFAs for each variable at each time point to ensure measurement invariance over time for each construct. I also tested growth in the three variables. I then tested a second-order factor (say y) for all three constructs but did not include growth curves. I now want to test a structural model of x predicting change over time in y.
Q1. How do I include change over time in my second-order latent factor y?
Q2. Can I even test x predicting change over time in y or should I test x predicting change over time in each of the variables without the second-order factor?
Thank you for your response. I think I was unclear in my question. You are right that there is a single factor at each time point. My second order factor includes all the factors across waves i.e., a1,a2,a3,b1,b2,b3,c1,c2,c3. Is this unconventional? I realize that I am not accounting for time in this model which is why I am trying to understand whether there is any way of testing change over time in a second-order factor.
If you have a second-order factor using the three factors, it is like an intercept only growth model but with free time scores. An intercept only growth model would have all factors loadings fixed to one. I would think using a conventional multiple indicator growth model would be more interpretable. See multiple indicator growth in the Topic 3 or 4 course handouts and videos.