Estimation Methods for Nest Success: Bayesian Analysis for Nest Success Models

摘要

Nest success models accounted for both age-dependent first encounter211u001eprobabilities and age-dependent failure probabilities have received much 211u001eattention recently. Major problems in the classical method include the lack of 211u001eclosed form of the maximum likelihood estimates (MLE) and the instability of the 211u001eMLE. To estimate first encounter probabilities and failure probabilities 211u001esimultaneously, a Bayesian approach is proposed using a large class of priors. 211u001eThe computation can be done via Gibbs sampling procedures. Some latent variables 211u001eare introduced to simplify the full conditional distributions. Sampling from the 211u001eindividual encounter posterior distribution is done using simple transformation. 211u001eFor a general class priors, a rejection method is suggested to sample from 211u001eindividual distributions of failure probabilities. The methods are illustrated by 211u001eusing data from birds with known fate and under regular visits. Comparisons are 211u001emade between the MLE and Bayesian estimates. The results indicated that the 211u001eBayesian analysis provided reasonable estimates of nest success rates.