Analysis of Count Data with Structural Zeros: CTN0018 and CTN0019 Project Summary/Abstract This application addresses two statistical issues in substance abuse research (1) the analysis of a count response with structural zeros, and (2) analysis of treatment moderators under a new moderation framework. Both of these statistical issues are highly relevant to the analysis of differential treatment effectiveness of HIV sexual risk interventions as examined within two NIDA Clinical Trials Network (CTN) studies: CTN0018 ("HIV/STD Safer Sex Skills Groups For Men In Methadone Maintenance Or Drug-free Outpatient Treatment Programs") and CTN0019 ("HIV/STD Safer Sex Skills Groups For Men In Methadone Maintenance Or Drug-free Outpatient Treatment Programs"). Our general goals in this application are to (1) develop new statistical methods for analysis of count data with structural zeros, (2) develop new statistical methods of analysis of moderators, (3) create software for implementing the new statistical methods, and (4) test hypotheses about moderators of treatment effectiveness using pooled data from CTN0018 and CTN0019. Specifically, we will develop a distribution-free, varying-coefficient class of models for moderation analysis of count data in the presence of structural zeros and over dispersion. This new approach addresses fundamental flaws in current methods that contribute to inconsistent findings in substance abuse and related research. We will examine two primary variables (severity of current drug use and engagement in sex under the influence of drugs/alcohol), and several secondary variables, as moderators of the effectiveness of a five session gender-specific HIV sexual skill intervention relative to a one session "treatment as usual" standard HIV education intervention. We hypothesize that (1) the five session gender-specific skills training interventions used in CTN0018 and CTN0019 will have greater effectiveness, relative to the one session standard intervention, for patients with higher levels of current drug use of their primary drug of abuse (but equal effectiveness for lower levels of current drug use), and (2) the five session intervention will have greater effectiveness, relative to the one session intervention, for patients who engage in sex under the influence of drugs/alcohol (but equal effectiveness for patients who do not engage in sex under the influence). The pooled study database will also allow for sufficient statistical power for testing of gender and minority status as potential moderators of treatment effectiveness.
Analysis of Count Data with Structural Zeros: CTN0018 and CTN0019 Project Narrative This application develops new methods for addressing two statistical issues in substance abuse research: (1) the analysis of a count response with structural zeros, and (2) analysis of treatment moderators under a new moderation framework. Both of these statistical issues are highly relevant to the analysis of differential treatment effectiveness of HIV sexual risk interventions. The new methods will have broad applicability for the accurate analysis of data from studies of substance abuse treatments and studies of HIV risk behavior, as well as other domains of research, and will help us understand whether HIV sexual risk behavior interventions work equally well for patients with different characteristics, allowing for such interventions to be efficiently implemented in clinical practice.
|Crits-Christoph, Paul; Gallop, Robert; Sadicario, Jaclyn S et al. (2014) Predictors and moderators of outcomes of HIV/STD sex risk reduction interventions in substance abuse treatment programs: a pooled analysis of two randomized controlled trials. Subst Abuse Treat Prev Policy 9:3|
|Tang, Wan; He, Hua; Gunzler, Douglas (2012) Kernel Smoothing Density Estimation when Group Membership is Subject to Missing. J Stat Plan Inference 142:685-694|
|Tang, Wan; Zuo, Guoxin; He, Hua (2011) Double-smoothing for Varying Coefficient Models. J Nonparametr Stat 23:917-926|