Bayesian analysis of lamb survival using Monte Carlo numerical integration with importance sampling

Area of Science:

• Animal Science
• Statistical Modeling
• Bayesian Inference

Background:

• Lamb survival from birth to weaning is a critical factor in sheep production.
• Understanding factors influencing lamb survival is essential for optimizing flock management.
• Previous studies have utilized various statistical approaches to analyze survival data.

Purpose of the Study:

• To compare approximate and exact Bayesian analyses of lamb survival data.
• To assess the impact of different statistical methodologies on survival probability estimations.
• To identify key factors affecting lamb survival in Rambouillet sheep.

Main Methods:

• Conducted approximate and exact Bayesian analyses on 2,554 Rambouillet lambs.
• Employed asymptotic normal approximation and Monte Carlo numerical integration with importance sampling.
• Utilized a linear logistic model to evaluate effects of year, dam age, sex, and birth type; also performed least squares analysis.

Main Results:

• Bayesian and non-Bayesian analyses produced virtually identical results for survival probability.
• Asymptotic normal approximations to posterior distributions were excellent, indicating informative likelihood functions.
• Lamb survival rates were higher for four-year-old ewes, similar for both sexes, and 10% higher for single births compared to multiple births.

Conclusions:

• Both Bayesian and frequentist approaches are effective for analyzing lamb survival data.
• Dam age and birth type are significant predictors of lamb survival.
• The findings provide valuable insights for sheep breeding and management strategies to improve lamb survival rates.