bayesZIB

bayesZIB models zero-inflated dichotomous (Bernoulli) data using Bayesian inference to distinguish structural and non-structural zeros for improved estimation.

Key Features:

• Bayesian framework: Implements a Bayesian framework tailored for zero-inflated Bernoulli (dichotomous) data.
• Zero-source separation: Explicitly models and distinguishes structural zeros (inherent absences within a subpopulation) from non-structural zeros (zeros due to sampling variability).
• Validation: Performance evaluated through comprehensive simulation studies measuring accuracy in identifying and modeling the dual sources of zeros.
• Implementation: Implemented as an R package.

Scientific Applications:

• Occupational health — sickness presenteeism: Separates individuals inherently not at risk from those with zero outcomes due to chance to avoid conflating general health status with the condition of interest.
• Zero-inflated dichotomous outcome analysis: Applicable to research contexts with dichotomous outcomes and excess zeros to reduce bias in estimation and inference.

Methodology:

Employs a novel Bayesian approach tailored to zero-inflated Bernoulli data and is evaluated using comprehensive simulation studies to assess identification and modeling of structural versus non-structural zeros.