Software
R packages developed with collaborators to implement our methods for causal inference, mediation analysis, and randomized trials.
lmtp
Estimates causal effects of longitudinal modified treatment policies, with support for continuous, categorical, and binary exposures; survival outcomes with competing risks; and flexible machine learning estimation.
Authors: Nicholas Williams and Iván Díaz.
crumble
Implements modern causal mediation analysis using Riesz learning. Supports several definitions of direct and indirect effects, intermediate confounding, and high-dimensional mediators and confounders.
Authors: Nicholas Williams and Iván Díaz.
medoutcon
Provides efficient natural and interventional mediation estimators, including settings with intermediate confounding and two-phase sampling.
Authors: Nima Hejazi, Iván Díaz, and Kara Rudolph.
medshift
Estimates direct and indirect effects of stochastic interventions using nonparametric mediation methods.
Authors: Nima Hejazi and Iván Díaz.
adjrct
Implements model-robust covariate adjustment for randomized trials with survival or ordinal outcomes, incorporating machine learning for efficiency gains.
Authors: Iván Díaz and Nicholas Williams.
Additional Team Software
lcm and lcmmtp
Implement mediation analysis with time-varying mediators and treatments, with support for high-dimensional confounders, categorical mediators, and Super Learner ensemble estimation.