Meaghan Winder (Dept. of Mathematical Sciences, MSU)

09/28/2023 - PhD Proposal

Abstract: As ecological questions become more complex, and statistical methods are developed around them, there is an increasing demand for making inference about quantities that are not directly observed; in order to estimate these latent quantities, scientists can utilize a hierarchical sampling design. For example, with occupancy data, sites are each sampled multiple times. During these repeated samples, it is possible that the species of interest is not detected, but is still present at the site. The hierarchical sampling design allows users to gain some understanding of the detectability of the species, and as a result, allows them to estimate the probability that a site is occupied. Here, we will discuss the first two chapters of my dissertation: 1) modeling the detectability of invasive mussels in water bodies across the western United States in order to make sampling recommendations for early detection and 2) a review of dynamic occupancy models and the implications of incomplete sampling on various dynamic occupancy model parameterizations. We will conclude with a discussion of the proposed third chapter of my dissertation related to Bayesian variable selection for hierarchical non-Gaussian latent variable models.