Software
Here is a list of R packages developed by Nick Williams implementing some of the methods that our team have developed:lmtp
R package implementing general estimators for the generalized g-computation formula. Some features include:
- Estimation of causal effects for static interventions and modified treatment policies in longitudinal data
- Support for continuous, categorical, and binary exposures
- Integration with Super Learner ensemble learners for flexible machine learning estimation of nuisance parameters
- Handles right-censored data
adjrct
R package implementing various estimators for covariate adjustment in randomized studies. Some features include:
- Estimation of covariate adjusted effects for randomized trials, with efficiency guarantees
- Estimation of effects for ordinal and time-to-event endpoints
- Incorporation of machine learning for maximum efficiency gains
- Handles right-censored time-to-event endpoints
crumble
R package implementing general estimators for mediation analysis and path analysis:
- Estimation of multiple effects in the literature, such as natural direct and indirect effects, randomized interventional effects, recanting twin effects, separable effects, etc.
- Incorporates high-dimensional (e.g., omics) mediators and confounders
- Deep learning estimation of Riesz representers
lcm and lcmmtp
R package implementing general estimators for mediation analysis with time-varying mediators and treatments:
- Incorporates high-dimensional confounders (but only categorical mediators)
- Integration with Super Learner ensemble learners for flexible machine learning estimation of nuisance parameters