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

Area of Science:

• Biostatistics
• Public Health
• Epidemiology

Background:

• Sexual behavioral outcomes often yield count data with excess zeros and overdispersion.
• Standard Poisson log-linear models are inadequate for such data characteristics.
• Existing alternative models are underutilized in sex health research, particularly HIV prevention studies.

Purpose of the Study:

• To discuss methods for analyzing count outcomes with excess zeros and overdispersion.
• To introduce appropriate model-fit indices for comparing statistical models.
• To provide guidance for sound statistical analyses in HIV prevention and related research.

Main Methods:

• Review of statistical modeling techniques for zero-inflated and overdispersed count data.
• Application of methods to real-world data from an HIV prevention intervention study.
• Evaluation and comparison of competing statistical models using model-fit indices.

Main Results:

• Demonstration of challenges posed by excess zeros and overdispersion in sexual behavior data.
• Identification of appropriate statistical models and model-fit indices for these data types.
• Illustrative analysis using data from an HIV prevention intervention.

Conclusions:

• Accurate modeling of sexual behavior count data requires specialized statistical approaches beyond standard Poisson models.
• Effective use of advanced statistical methods and model-fit indices can improve research quality in sex health.
• This work facilitates better statistical analysis and research in HIV prevention and related behavioral studies.