Baltic Sea Peracarids Diurnal Activity Data : Issues On Multiple Imputation in Overdispersed Poisson Regression Model and Model Selection

Abstract

Abtsrcat

In this paper, we consider the diurnal (day time) activities of peracarisdis in Baltic sea. The peracaridis is a large group of malacostracan crustaceans, having members in marine, freshwater, and terrestrial habitats. They are defined by the possession of a single pair of maxillipeds (rarely 2-3), of mandibles with an articulated accessory process between the molar and incisor teeth in the adults (called the lacinia mobilis), and of a carapace which is often reduced in size and is not fused with the posterior thoracic somites. The experimental data is given in Janson and Kallander (1968). In this study, the animals are caught by sweeping a net over a known distance and fixed time intervals. The total number of caught along with various covariates are recoded. The data is missing some responses and some covariates. In view of that we develop Bayesian approaches for modeling overdispersed Poisson data in the context of regression problem in presence of missing data. We handle the missing data problems through various multiple imputation schemes and take a Bayesian route for model selection to determine the appropriate scheme for the given data.