Questions on Group-Based Trajectory Modeling (GBTM) for Categorical Outcomes

[Black Hat]

I am working on a project involving group-based trajectory modeling (GBTM) and have several questions I hope you can help with. My dataset tracks treatment adherence (a binary variable: yes/no) over several years (2012-2016). Here are my queries:

1. Can GBTM be applied to categorical outcomes (treatment adherence: yes/no)?
2. Which variables should be considered to identify trajectories? Should I use just the time variable (year) and the outcome (treatment adherence), or do I need to adjust for other covariates?
3. What should I do if the odds of correct classification (OCC) for several groups is less than 5 and the average posterior probability is below 0.7?
4. Why am I getting random BIC values for increasing degrees? Usually, the BIC decreases with increasing degrees, but in my case, for 1 group it might be -7800, for 2 groups it might be -7600, and for 3 groups it might be -7700.

I am a layman without much knowledge on statistics or stata. Please, answer to the questions in digestible way for me. Thank you in advance for your assistance. Your insights will be greatly appreciated.

[Nick Cox]

Please read and act on https://www.statalist.org/forums/help and https://www.statalist.org/forums/help#adviceextras

If Black Hat is your real name, so be it; otherwise please note our explicit request to use full real names, say given name and family name.

Otherwise this kind of questionnaire is unlikely to be successful if only because you don't give any real details of your data or what Stata commands you used (if any) or any precise results. It's written as if you somehow expect that we are looking at your computer and know what data you have and see what you are seeing.

There is no context here. If this is an assignment or a dissertation or doctoral project, you should be seeking whatever support is allowed from teaching assistants, supervisors, or committee members. If you are say a medic working with some data, you surely need a collaborator or consultant to give support.

Cross-posted at https://stats.stackexchange.com/ques...rical-outcomes and https://www.reddit.com/r/rstats/comm...modeling_gbtm/

We ask that you give explicit mention of any cross-postings.