On comparison of mixture models for closed population capture-recapture studies

Area of Science:

• Ecology
• Biostatistics
• Population Dynamics

Background:

• Capture-recapture studies often face challenges with individual heterogeneity.
• Mixture models are a common approach to account for such variations.
• Existing models like the two-point mixture and beta-binomial have debated advantages and limitations.

Purpose of the Study:

• To compare the performance of the two-point mixture model and the beta-binomial model in capture-recapture studies.
• To investigate the sources of bias, specifically approximation and estimation bias, in these models.
• To analyze the impact of boundary problems on population size estimation.

Main Methods:

• Decomposition of total bias into intrinsic, approximation, and estimation bias.
• Exclusion of intrinsic bias for a focused comparison of estimators.
• Investigation of boundary problems and their effects on model performance.
• Application of models to real-world epidemiological and ecological data.

Main Results:

• The study quantifies approximation bias, estimation bias, and variance for both models.
• Boundary problems are identified and their influence on population estimates is assessed.
• Comparative analysis of model performance using empirical datasets.

Conclusions:

• Mixture models are essential for handling heterogeneity in capture-recapture data.
• The study provides a framework for comparing estimators by excluding intrinsic bias.
• Understanding bias and boundary issues is crucial for accurate population size estimation in ecological and epidemiological contexts.