While causality is a highly explored area of research in biostatistics and epidemiology, it has not been explored extensively for ecological data. Causal approaches have the potential to establish or uncover causal links and estimate their effects in large ecosystem networks. However, many causal methods require separability, which is not possible to establish in nonlinear systems common in nature. Causal methods have the potential to identify causes behind population fluctuations linked to climate change and plan interventions to reduce the impacts of climate change on community dynamics. I am establishing the use of causal methodologies for ecological questions and improving these methodologies for detecting and estimating causal influences in dynamic, nonlinear systems. I encourage wider utilization of causal reasoning and causal methodologies for ecological research that seeks to establish and estimate causal relationships for developing ecological theory or making management and policy decisions.

My research focuses on developing and applying robust and efficient nonparametric statistical procedures for nonlinear modeling of climate change effects on groundfish communities in the northern Pacific Ocean. Working with non- and semi-parametric models to explore the effects of climate changes on populations has highlighted the inadequacy of current methods to appropriately calculate effect size measures for predictors in high-dimensional model structures. As model structures become complex in order to accommodate more ecological interactions, accurately measuring the size of such effects becomes equally convoluted. I am therefore interested in creating robust effect size measures suitable for such model structures to expand the usability of these models to scientists who require robust inferences for a wide range of applications.

Studies of multiple paternity in mammals and other animal species generally report proportion of multiple paternity among litters, mean litter or clutch sizes, and mean number of sires per litter or clutch. I collaborate with a population ecologist and a statistician to estimate a null model for multiple paternity across a variety of animals using Bayesian regression models. We show how these variables can be used to produce an estimate of the probability of reproductive success for a male that has mated with a female. This estimate of male success is more closely aligned to the intensity of sexual selection and is more informative about the mating system that alternative measures, like the proportion of litters with multiple paternity or the number of sires. The probability of success for a mated male can be measured both theoretically and empirically, and gives an estimate of a male’s “confidence of paternity” upon mating.