(almost) everything I did in grad school. Note that I removed most of the datasets and some of the rendered pdfs because of size constraints. Most of this is in R, with a mix of Python and MATLAB as well.
A class on algorithmic methods applied to stock market trading. This repo contains homework. Our final project, which used gradient boosting and fundamental stock selection can be viewed here. We won our in-class trading competition with this approach!
Duke Statistical Science is a very Bayesian-heavy department, so we spent a significant amount of time learning various Bayesian methods including MCMC.
This includes homeowkr from classical inference, as well as the two most difficult exams I've ever taken.
Study of geospatial models and time series, including AR, ARMA, ARIMA, CAR, SAR, and Gaussian Process models.
A course that combined frequentist and Bayesian approaches to building hierarchical models. There were two projects, which are in separate repositories:
- Case Study 1 An exploration of self-reported drug pricing data.
- Case Study 2 An examination of NC election data. You can view my presentation on YouTube here.
A class on the statistical and probabilistic underpinnings of machine learning.
- Our final project, applying machine learning methods to predict fish biomass levels.
A course in R.
The most difficult class I've ever taken. I learned a tremendous amount, though! I completed two projects which are in separate repos:
- Project 1: Improved plots from a published paper, and then applied various machine learning methods to better understand the health of the trees.
- Project 2: A classification task to predict the presence of clouds vs. ice in satellite images of the arctic.
A class examining different sampling techniques, study and survey design, and a brief introduction to causal inference.