Doubly Inflated and Truncated Models with Application in Self-Reported Drug Usage Among Rural African-American Cocaine Users

• Activity Number: 552
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Songthip Ounpraseuth*+ and Horace J Spencer and Jeff Thostenson and Brenda M. Booth and Katharine E Stewart
• Companies: UAMS and UAMS and UAMS and UAMS and UAMS
• Keywords: Doubly inflated model ; Doubly truncated model ; Respondent driven sampling ; rural population

Abstract

The focus of this study is to illustrate the differences among various double inflated models evaluating the impact of two community-based interventions on self-reported drug usage in rural African-American cocaine users in Arkansas.

Respondent-driven sampling (RDS) was used to recruit participants in two rural, under-resourced, predominantly African-American counties in the Arkansas Delta region. Neither intervention group focused on drug use but both included information regarding risks associated with usage. Doubly inflated Poisson, doubly truncated Poisson, and right truncated zero-inflated Poisson models were each fit to the data on the intent-to-treat basis. The dependent variable was the number of days in the past 30 days in which participants used crack and/or powdered cocaine. Approaches used for examining goodness-of-fit include -2LL, the AICC and the BIC.

A total 251 participants were enrolled. Our results illustrate the potential advantages of an inflated and truncated mixture approach to handling extreme stacking of data at endpoints in drug usage outcomes in terms of model fit and a more accurate interpretation.