Modeling Count Outcomes from HIV Risk Reduction Interventions: A Comparison of Competing Statistical Models for Count Responses

Department of Biostatistics and Computational Biology, Box 630, University of Rochester, 265 Crittenden Boulevard, Rochester, NY 14642, USA.
AIDS research and treatment 03/2012; 2012(2):593569. DOI: 10.1155/2012/593569
Source: PubMed


Modeling count data from sexual behavioral outcomes involves many challenges, especially when the data exhibit a preponderance of zeros and overdispersion. In particular, the popular Poisson log-linear model is not appropriate for modeling such outcomes. Although alternatives exist for addressing both issues, they are not widely and effectively used in sex health research, especially in HIV prevention intervention and related studies. In this paper, we discuss how to analyze count outcomes distributed with excess of zeros and overdispersion and introduce appropriate model-fit indices for comparing the performance of competing models, using data from a real study on HIV prevention intervention. The in-depth look at these common issues arising from studies involving behavioral outcomes will promote sound statistical analyses and facilitate research in this and other related areas.

Download full-text


Available from: Yinglin Xia
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In the case of sensitive questions such as number of alcoholics known, majority of respondents might give an answer of zero. Poisson regression model (P) is the standard tool to analyze count data. However, P provides poor fit in the case of zero inflated counts, when over-dispersion exists. Therefore, the questions to be addressed are to compare performance of alternative count regression models; and to investigate whether characteristics of respondents affect their responses. A total of 700 participants were asked about number of people they know in hidden groups; alcoholics, methadone users, and Female Sex Workers (FSW). Five regression models were fitted to these outcomes: Logistic, P, Negative Binomial (NB), Zero Inflated Poisson (ZIP), and Zero Inflated Negative Binomial (ZINB). Models were compared in terms of Likelihood Ratio Test (LRT), Vuong, AIC and Sum Square of Error (SSE). Percentages of zero were 35% for number of alcoholics, 50% for methadone users, and 65% for FSWs. ZINB provided the best fit for alcoholics, and NB provided the best fit for other outcomes. In addition, we noticed that young respondents, male and those with low education were more likely to know or reveal sensitive information. Although P is the first choice for modeling of count data in many cases, it seems because of over-dispersion of zero inflated counts in the case of sensitive questions, other models, specifying NB and ZINB, might have better goodness of fit.
    Full-text · Article · Sep 2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We describe a general framework for modeling and stochastic simulation of epidemics in realistic dynamic social networks, which incorporates heterogeneity in the types of individuals, types of interconnecting risk-bearing relationships, and types of pathogens transmitted across them. Dynamism is supported through arrival and departure processes, continuous restructuring of risk relationships, and changes to pathogen infectiousness, as mandated by natural history; dynamism is regulated through constraints on the local agency of individual nodes and their risk behaviors, while simulation trajectories are validated using system-wide metrics. To illustrate its utility, we present a case study that applies the proposed framework towards a simulation of HIV in artificial networks of intravenous drug users (IDUs) modeled using data collected in the Social Factors for HIV Risk survey.
    Full-text · Article · Apr 2014 · Simulation
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In psychosocial and behavioral studies count outcomes recording the frequencies of the occurrence of some health or behavior outcomes (such as the number of unprotected sexual behaviors during a period of time) often contain a preponderance of zeroes because of the presence of 'structural zeroes' that occur when some subjects are not at risk for the behavior of interest. Unlike random zeroes (responses that can be greater than zero, but are zero due to sampling variability), structural zeroes are usually very different, both statistically and clinically. False interpretations of results and study findings may result if differences in the two types of zeroes are ignored. However, in practice, the status of the structural zeroes is often not observed and this latent nature complicates the data analysis. In this article, we focus on one model, the zero-inflated Poisson (ZIP) regression model that is commonly used to address zero-inflated data. We first give a brief overview of the issues of structural zeroes and the ZIP model. We then given an illustration of ZIP with data from a study on HIV-risk sexual behaviors among adolescent girls. Sample codes in SAS and Stata are also included to help perform and explain ZIP analyses.
    Full-text · Article · Aug 2014 · Shanghai Archives of Psychiatry
Show more