Bayesian inference on age-specific survival for censored and truncated data

Area of Science:

• Ecology
• Population Biology
• Statistical Modeling

Background:

• Estimating age-specific survival and mortality in vertebrates traditionally requires known ages, limiting studies with incomplete records.
• Existing capture-recapture and capture-recovery models face challenges with large proportions of unknown ages and insufficient study durations.

Purpose of the Study:

• To present a novel hierarchical Bayesian model for capture-recapture/recovery (CRR) data with a high proportion of unknown birth and death times.
• To reconstruct individual birth and death times and estimate population-level demographic rates using parametric survival functions.

Main Methods:

• Developed a Bayesian hierarchical model for CRR datasets with missing age information.
• Simulated CRR datasets with varying parameters to compare the new model against traditional CRR models.
• Applied the model to a long-term Soay sheep CRR dataset.

Main Results:

• The proposed model outperforms traditional CRR models with low sample sizes.
• The model accurately estimated ages at death for Soay sheep (average error 0.94 years) and identified sex-specific mortality differences.
• Model performance is sensitive to prior choice and study duration/recapture probability, with overestimating priors performing better in challenging conditions.

Conclusions:

• The developed model provides reliable estimates of demographic parameters and individual birth/death times from incomplete ecological data.
• This approach enhances the analysis of sparse datasets common in wildlife research.
• Combines improved sampling with advanced statistical modeling to overcome fieldwork limitations.