Fixed and Random Effects Models and Multistage Estimation Procedures for Statistical Population Reconstructions
Abstract
Age-at-harvest data are routinely collected as part of game-management programs. These data represent a wealth of information regarding demographic processes and trends in wildlife abundance. Use of wildlife age-at-harvest data has blossomed only relatively recently in the literature despite its frequent collection by game management agencies. Statistical models exist for such data, but are limited in their facility, owing to restrictive assumptions regarding constancy of demographic processes, unsuitability of models for the type of data collected, or computing difficulty in fitting models. Current models cannot accommodate the presence of process error (natural variation in demographic processes), or separate this error from sampling error (measurement error that is present whenever the full population cannot be sampled). I develop and examine a set of statistical models for demographic processes using primarily age-at-harvest data that can be used to estimate survival probability, harvest vulnerability, and recruitment, as well as process error associated with these entities. I conduct thorough simulation studies of these models, and assess them with respect to their ability to accurately and precisely reconstruct abundance. Studies are conducted for fully-aged big game harvest, pooled age-class big-game harvest, and small-game harvest. Results indicate that a mixed-effects model which incorporates random effects in the processes of natural mortality and harvest probability, as well as a likelihood conditional on total cohort capture along with a Horvitz-Thompson abundance estimator outperform other models, and are recommended for use.