Hi, there.

I am a Harvard Data Science Initiative Postdoctoral Fellow working with Francesca Dominici and Tyler VanderWeele at the Harvard TH Chan School of Public Health. I received a PhD in Biology and MS in Statistics at Auburn University.

My research interests broadly cover the area of statistical ecology, particularly developing new techniques to quantify climate change effects on populations and identifying causation in ecological data. I also aim to improve statistical methods for high-dimensional and spatio-temporal data.


An overview of my research: statistical ecology

Developer Image
Population dynamics of Amboseli elephants (shown) adjusted for effects of rainfall on birth rates.
Developer Image
Coefficient functions of a single-index varying coefficients regression model estimated by RSGLASSO under CN(0.95) error distribution.
Developer Image
Bayesian MCMC estimated frequency of multiple paternity across litter sizes for mammalian species (red points). The solid red line is the predicted frequency of multiple paternity using a zero-truncated binomial distribution.
Developer Image
Spatial and temporal effects of plant productivity and herd condition on juvenile body mass of Eurasian reindeer (shown during Sami racing in Tromsø) in Norway.
Developer Image
Modeling predator-prey dynamics of adult groundfish in the Gulf of Alaska. Shown is the Pacific halibut, an apex predator in the northeastern Pacific Ocean.
Developer Image
Maximum correlation of cross-mapped versus observed values as a function of time series length for comparisons between environment, halibut, and cod (processes A, as columns) and sablfish CPUE (process B). Significant forcing of A on B is represented as red horizontal bars, while significant forcing of B on A are blue vertical bars. Non-significant relationships are indicated as dark grey horizontal bars (A→B) and vertical bars (B→A). Spearman’s rank correlation coefficient is represented as open circles. Empty plots indicate that there were insufficient observations from a complete time series for the multispatial CCM algorithm to begin. Plots with only Spearman’s indicate that the predictive ability of one or both processes did not significantly decrease with increasing time distance, so CCM was not performed.