Estimation and inference with complex count dataudfrom fisheries surveys, including over-dispersion,udmany nuisance parameters, and correlation

摘要

We study methods to estimate regression and variance parameters for over-dispersedudand correlated count data from highly stratified surveys. A challenge with such dataudis the large number of nuisance parameters which leads to computational issues andudbiased statistical inferences. We develop a profile generalized estimating equationud(GEE) method that is more computationally efficient and compare it to marginaludmaximum likelihood (MLE) and restricted MLE (REML) methods. We use REMLudto address bias and inaccurate confidence intervals because of many nuisance parameters.udThe marginal MLE and REML approaches involve intractable integrals and weudused a new R package that is designed for estimating complex nonlinear models thatudmay include random effects. We conduct simulation analyses and conclude that theudREML method is the better approach among the three methods we investigate.udOur applications involve counts of fish catches from highly-stratified research surveys.udIn the first application, we estimate the day and night (diel) effect for threeudspecies from bottom trawl research surveys. In the second application, we estimateudthe diel and vessel effects of two different snow crab surveys.