Hi. I have longitudinal data for growth modeling. There are 485 persons who were asked questions before and after quitting smoking. Outcome variable is kind of continuous, which is how much they want to smoke. I would like to see the trajectories before and after quitting. I am thinking four latent variables, intercepts and slopes before and after. Do you think these data are fit to the Ex. 6.13. in the Mplus user's guide(Growth model for two parallel processes for continuous outcomes...)? If it is right, I have one more question. My data format is typical longitudinal format with SPSS, how can I change them into multivariate format for latent growth modeling? I would appreciate it if you just say "check this website" or "see this book" or something. Thank you in advance.
bmuthen posted on Saturday, October 01, 2005 - 2:16 pm
This is an example of sequential, not parallel processes. It can be seen as a form of "piecewise" growth modeling (see e.g. the Raudenbush-Bryk book). But you have the complication that different people quit at different times, so the point of changing from one "piece" to the next varies across people. This leads to a much more complex model, which has been studied in the literature as "changepoint" modeling. Such changepoint modeling cannot currently be done in Mplus. A similar approach can perhaps be achieved by using Type = Mixture and let different latent classes correspond to different quitting times.
Regarding your data question, note that Mplus can do growth modeling with data in the "long" form that it sounds like you have it in for SPSS. This would be using twolevel growth modeling using Type=Twolevel. If you want to do growth modeling as a multivariate single-level analysis with data in "wide" form as described in the User's Guide, you will have to rearrange the data yourself with different data columns corresponding to the outcome at different time points (Mplus will in the next version have a way to automatically go back and forth between long and wide data forms).