Bernoulli Ballot Polling: A Manifest Improvement for Risk-Limiting Audits
Published:
Ballot polling RLAs using Bernoulli sampling, rather than simple random sampling, can bypass current logistical challenges.
Download here
Research
Risk-Limiting Audits by Stratified Union-Intersection Tests of Elections (SUITE)
Published:
We’ve developed a new statistical method for risk-limiting post-election audits of stratified samples of ballots.
Download here
The Risk Limit of Bayesian Audits
Published:
I presented this work at the E-VOTE-ID 2018 PhD Colloquium.
Download here
Random Problems with R
Published:
Faulty algorithms in R’s random sampling functions.
Download here
Next Steps for the Colorado Risk-Limiting Audit (CORLA) Program
Published:
We proposed a strategy for Colorado to conduct risk-limiting post-election audits of contests spanning jurisdictions that use heterogeneous voting systems.
Download here
The Practice of Reproducible Research: Case Studies and Lessons from the Data-Intensive Sciences
Published:
I contributed two case studies of my data science workflow.
Download here
An Empirical Comparison of Parametric and Permutation Tests for Regression Analysis of Randomized Experiments
Published:
We compared traditional ANCOVA to permutation approaches in the analysis of randomized experiments.
Download here
Nonparametric Risk Attribution for Factor Models of Portfolios
Published:
We explain a two-step process for partitioning the risk of projected returns into contributions from latent factors using nonparametric regression methods.
Simple Random Sampling: Not So Simple
Published:
We propose several best practices for researchers using PRNGs for simulations, including the wide adoption of hash function based PRNGs.
permute: a Python Package for Permutation Methods
Published:
permute is a Python package for permutation testing.
Download here
Student Evaluations of Teaching (Mostly) Do Not Measure Teaching Effectiveness
Published:
We showed that student evaluations of teaching are biased against female instructors, and it is not possible to adjust for the bias due to its dependence on many other factors.
Download here