Statistics Colloquium

Title:  A Characterization of Shark Behavior using GPS Tracking Data

Abstract: Animal behavior poses many challenges for statistical modelling. Biologists argue that this behavior can be explained (in part) by
animal movements over time. This enables researchers to quantify
the interpretation of animal movements. It is of particular interest
to evaluate shark interactions. We devise a variety of statistical
models for this purpose. The first employs extended Kalman Filters to describe the movements of each shark separately. We then devise a Bayesian methodology to compare and contrast the movements of the different sharks. These include both direct comparisons of hyperparameters and Bayesian p-value like constructions. The second employs BART and Dirichlet Processes to directly describe the various shark states through a logistic-regression model structure.