About me
I am a doctoral student in the Archie Carr Center for Sea Turtle Research at the University of Florida. I am studying life-history theory and the evolution of aging in Professor Karen Bjorndal’s lab. I hold a M.Sc. in wildlife conservation from Trinity College Dublin.
I currently research senescence in testudines, and I am particularly interested in the intersection between life-history strategy and rates of aging. I approach my research through a statistical framework. While, for the most part, my focus has been on sea turtles, I am deeply interested in how all these interacting factors influence population dynamics. I feel that by understanding the underlying mathematics of biological systems, you gain a deeper insight into the nuts and bolts of how these systems function.
I have worked in various roles on multiple projects, both governmental and non-governmental, around the world. These include ARCHELON, the Sea Turtle Protection Society of Greece, Reef HQ, The Great Barrier Reef Aquarium, and the Sea Turtle Conservancy.
Senescence in sea turtles
Here we present the first paper researching aging dynamics in testudines including sea turtles.
Statistical bias in parameter estimation
Hierarchical models (or mixed-effects models) are the bread and butter of statistical ecology and genetics. We all use and have used them. This reliance is unlikely to change because of their broad applicability, but commonly used packages like glmmTMB have an Achilles heel! We show this fact in explicit detail in this paper. While these results may seem relegated to dark statistical corners – fated as nothing other than intellectual discussions or disagreements among competing methods that ultimately produce the same result – this most certainly is not the case! The methods and results in this paper have real-world implications for how we all do science. For instance, this issue could be the difference between incorrectly concluding that a treatment or a biotic factor (like temperature) has a positive or negative effect on a response, such as population size and persistence. And because there are no diagnostics to show the model is wrong, we are susceptible to perpetuating false and irreplicable claims. In this paper, we show there are applicable diagnostics to determine if methods using the Laplace approximation are appropriate, using ML estimation via MCMC and data cloning.
Affiliations
