Software
R packages I maintain or contribute to. Most implement methods from my papers; a few are infrastructure for collaborative projects.
sensemakr
Sensitivity analysis tools for regression results to unobserved confounding. Implements the methods in Cinelli & Hazlett (2020, JRSS B): robustness value, bias-adjusted estimates, contour plots, extreme-scenario bounds. Maintained by Carlos Cinelli; see his documentation site for illustrations and demos.
Install in R with install.packages("sensemakr"), or in Stata with ssc install sensemakr.
KBAL
Kernel balancing for estimating causal effects and population-level quantities without imposing a parametric outcome model. Balances on the means of features in a reproducing-kernel Hilbert space — a nonparametric analog to entropy balancing that degrades gracefully under misspecification.
Includes the kpop() wrapper for survey reweighting, implementing the method in Hartman, Hazlett & Sterbenz (2024), Kpop: A kernel balancing approach for reducing specification assumptions in survey weighting, JRSS A.
Install with install.packages("kbal").
gpss
Gaussian Processes for Social Science. Non-parametric regression with principled uncertainty quantification — the posterior reflects the lesser knowledge we have at the edge of or beyond the observed data, where other approaches become highly model-dependent. Reduces user-chosen hyperparameters from three to zero, with convenience functions for regression discontinuity (gp_rdd()), interrupted time-series (gp_its()), and general GP fitting.
Implements the method in Cho, Kim & Hazlett (2026), Inference at the Data’s Edge, Political Analysis.
Install with install.packages("gpss").
scqe
Stability Controlled Quasi-Experiment. Bounds causal effects when only one pre-treatment and one post-treatment period are observed, using stability-of-trend assumptions instead of parallel trends.
A companion browser-based app lets you run SCQE on aggregate data with no R setup — useful for quick exploration and teaching.
Install with install.packages("scqe").
KRLS
Kernel Regularized Least Squares. A flexible nonparametric regression that relaxes the linearity assumption while staying interpretable: coefficients have the familiar marginal-effect interpretation, and results degrade gracefully under misspecification. Widely used as a machine-learning alternative to OLS in political science and policy analysis.
Implements the method in Hainmueller & Hazlett (2014), Political Analysis. The Stata port is described in Ferwerda, Hainmueller & Hazlett (2015), Journal of Statistical Software.
Install in R with install.packages("KRLS"), or in Stata with ssc install krls.
CBPS
Covariate Balancing Propensity Score. Contributor to the R package implementing propensity score estimation that simultaneously models treatment and balances covariates — including extensions to continuous treatments.
Install with install.packages("CBPS").
EarlyWarningProject
Statistical pipeline for the Early Warning Project’s annual forecasts of mass-atrocity risk, in partnership with the US Holocaust Memorial Museum.
Smaller tools and replication code for specific papers are on my GitHub profile.