Item Response Theory for Sequential Data in Drug Therapy

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Research explicating the types of family processes connected to the onset, course, and cessation of drug abuse among adolescents is critical for ongoing theory and intervention development for this population. The majority of what is known regarding family processes related to adolescent drug abuse is based on self-report methods of family assessment. Recently, however, interest and advances in observational methods of family research have surged considerably. Family researchers are increasingly utilizing observational assessment strategies to capture subtle and dynamic theory-based family communication processes which are not well measured using self-report methods. Unfortunately, prevailing analytic approaches for observationally coded data are limited in the capacity to directly model theory-based processes underlying observed sequences of communication. Consequently, little direct information is provided to inform theory and intervention development. In recognition of the need for advances in data analytic strategies for observationally coded data on family interaction, the proposed project aims to develop a novel application of item response theory (IRT) which can be applied to microcoded data from observational family assessments. IRT is well suited for providing fine-grained information about specific patterns of behavior characterizing underlying latent traits or processes. Compared to current methods of sequential data analysis, all of which are grounded in classical test theory, an IRT analytic approach can more fully capitalize on the richness of observational data for conducting theory-driven research on processes underlying observed sequences of communication. Although IRT has historically been utilized for psychometric analyses of traditional test items, new developments in IRT software have greatly expanded its range of potential applications in psychological and behavioral research. The current study will develop an application of IRT implemented within the very flexible generalized linear mixed modeling (GLMM) framework. The GLMM formulation enables adaptations of IRT to fit the unique contingencies of observational data, including random variation in the number and types of observed communication sequences between families. The specific aims of the proposed project are (a) formulate an IRT model capable of modeling observed sequences of family communication, (b) execute the IRT model within a nonlinear mixed modeling framework using observational data on patterns of communication in families of adolescents in treatment for substance abuse, (c) use the IRT model parameters to develop descriptive profiles of observed sequences of communication characterizing different levels of "family cohesion", and (d) use the IRT model parameters to develop profiles of "family cohesion" based on observed patterns of communication that are associated with levels of adolescent substance abuse.