Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | April Zeoli <zeoli@msu.edu> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: multistage survival data |
Date | Wed, 10 Mar 2010 13:27:56 -0500 |
Hello, I am working with longitudinal data which tracks the trajectory of people with mental health disorders who are released from jail. The events of interest are subsequent mental health treatment and subsequent incarceration, and many participants experience multiple incarcerations. The dataset contains dates of the events of interest, so I believe that a survival analysis is most appropriate to answer the question of whether subsequent mental health treatment delays the progression of a participant back to jail. At first I considered a recurrent event survival analysis, using Cox proportional hazards model, because many participants experienced multiple incarcerations. However, I struggled with how to model the mental health treatment variable. A dichotomous variable representing whether participants accessed mental health treatment post-jail omits potentially valuable information regarding whether the amount of time it takes to access treatment affects time to subsequent incarceration, or whether the type of mental health treatment accessed (information I also have) affects time to jail post-treatment. I am now exploring the use of a marginal survival analysis, modeling the data as multistage data, meaning that I am interested in numerous stages, or events (i.e., entry into study at time released from 1st jail, 1st mental health treatment, 2nd incarceration, 2nd mental health treatment, 3rd jail, etc.). I believe this would allow me to address the above questions. I would like to run a model that is a mix of conditional and marginal estimation, and include such conditional variables as gender and age. I discovered this type of model in "handbook of statistics 23 - Advances in Survival Analysis" edited by N. Balakrishnan and C.R.Rao, which states that proportional hazards models can be used for this purpose. The problem is that I do not know how to implement such a model in Stata 10.1 and have found no guidance as of yet. I am looking for guidance of any sort to help me with this analysis. Thank you for your consideration! April April M. Zeoli, PhD, MPH Assistant Professor School of Criminal Justice Michigan State University * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/