The comorbidity of substance use and eating disorders is emerging as a serious problem affecting adolescent girls (Foster & Richter, 2003). How this comorbidity develops is not well understood. Investigators wishing to study this phenomenon are faced with small samples, which create problems for some statistical techniques that could be tremendously useful. For example, latent transition analysis (LTA) has been used extensively to model the development of substance use (Collins & Wulgalter, 1992). LTA provides a promising framework for testing models that place development of eating disorders within the context of substance use. However, currently, LTA is not suitable for samples as small as those typically found in eating disorder research. The two aims of the proposed project are: (1) to develop and test a small sample version of LTA using a Bayesian approach, and (2) informed by the results from using Aim #1, investigate differential patterns of substance use among adolescent females with and without eating disorders. Products will include peer-reviewed scientific publications on the comorbidity of substance use and eating disorders, Bayesian methods in general, and on the new version of LTA for use in other drug use research applications.
