|
|
GMM trajectories to predict outcomes |
|
Message/Author |
|
|
Hi, I have longitudinal data on cigarette smoking for subjects over the course of the lifespan. What we would like to do is use GMM (or some variant) to identify different trajectories of cigarette smoking over the course of the lifespan. We would also like to know whether these trajectories predict outcomes (e.g., disease, mortality). My question is whether or not there exists a method in which these outcomes can be predicted simultaneously in a trajectory analysis, or whether we need to use a 2 stage approach (classify then analyze) where we identify classes of trajectories, assign individuals to trajectory groups based on posterior probabilities, and then determine whether there are group differences in disease and mortality rates (or time to these events) based on this classification. I like the idea of the classify then analyze approach, but the downside is that it doesn't take into account that the classification is imperfect. Your thoughts on this issue would be greatly appreciated. Thanks, Jen Rose |
|
|
Hi Jen, Example 8.6 from the user's guide presents what you are asking about. Also, you can practice and learn about this procedure by following along with one of Bengt's previous courses: http://www.gseis.ucla.edu/faculty/muthen/courses.htm I think assignment #6 addresses growth mixture modeling with a distal outcome. |
|
|
For a GMM example with a distal outcome, see also Muthén, B., Brown, C.H., Masyn, K., Jo, B., Khoo, S.T., Yang, C.C., Wang, C.P., Kellam, S., Carlin, J., & Liao, J. (2002). General growth mixture modeling for randomized preventive interventions. Biostatistics, 3, 459-475. which is posted on our web site. You are right that you don't want to classify/analyze because you will get biased estimated and distorted SEs. |
|
Back to top |
|
|