Nour Azhari posted on Wednesday, February 28, 2018 - 6:49 pm
I am trying to do mixture modeling on a dataset where the outcome variable is alcohol consumption per day during a treatment study. The treatment involves 5 week outpatient psychotherapy with a randomized medication infusion once on Week 2 for all participants. My problem is that each participant has a different length of treatment (because of missed appointments and rescheduling, some who dropped out earlier). How should I clean my data so that I have a meaningful interpretation once I run it in Mplus? And do all participants need to have the same number of time points?
Also, because the medication infusion is such a pivotal part of the treatment (from preliminary descriptives it seems that the day where participants received the medication, alcohol consumption decreased dramatically for many of them), I was wondering if there was a way I could arrange my data or model so that I could know that a specific time point in my model corresponds to the day of the infusion for all participants. In order to visualize the changes in a more meaningful way.
Muthén, B., Asparouhov, T., Hunter, A. & Leuchter, A. (2011). Growth modeling with non-ignorable dropout: Alternative analyses of the STAR*D antidepressant trial. Psychological Methods, 16, 17-33. Click here to view Mplus outputs used in this paper. download paper contact first author show abstract
This general analysis question is suitable for SEMNET.
It's hard to say which strategy to take without knowing more about your study - and we don't have time to get into such detailed communication.
It sounds to me that you have T=35 time points and at varying time points you randomize to different medications added to the psych therapy and that added treatment has a large and perhaps lasting effect and is of primary interest. If so, you could take the approach that we show in our Short Course topic 3, slides 157-159 where the 0/1 "status" variables correspond to your medication treatment.