Below are some R packages I have written or contributed to.

WeightIt: Weighting for Covariate Balance in Observational Studies

Noah Greifer

CRAN Page | GitHub Page | Website

{: .align-left} WeightIt implements a variety of weighting methods, including propensity score/inverse probability weighting, for achieving covariate balance in observational studies. Weights are available for binary, multi-category, continuous, and sequential treatments. Weight estimation methods include logistic regression, generalized boosted modeling, covariate balancing propensity score, entropy balancing, SuperLearner, BART, and more. Tools are available for assessing the distribution of weights and trimming weights if needed, and WeightIt objects are compatible with cobalt for balance assessment after weighting.{:style=“clear: left”}

MatchIt: Nonparametric Preprocessing for Parametric Causal Inference

Daniel Ho, Kosuke Imai, Gary King, Elizabeth Stuart, and Noah Greifer (maintainer)

CRAN Page | GitHub Page | Website

MatchIt implements a variety of matching methods, including propensity score matching, for achieving covariate balance in observational studies. A variety of methods of estimating propensity scores are available, including logistic regression, random forests, neural nets, BART, and covariate balancing propensity scores. Matching methods include nearest neighbor, optimal, full, genetic, exact, and coarsened exact matching and propensity score subclassification. A variety of options are available for customizing the method of matching. Extensive documentation has been written demonstrating how to perform matching, assess the quality of maytches, and estimate treatment effects after matching. Tools are available for assessing the quality of matches, and MatchIt objects are compatible with cobalt for balance assessment after matching. Though not one of the original authors, I performed a massive update of the package to version 4.0.0 and am the current maintainer.

cobalt: Covariate Balance Tables and Plots

Noah Greifer

CRAN Page | GitHub Page | Website

cobalt provides tools for assessing and reporting covariate balance in observational studies. Its strength is in being compatible with many other R packages, including MatchIt, WeightIt, twang, CBPS, and others, allowing for simple and unified balance assessment and ease in comparing the performance of several balancing methods. cobalt is extremely flexible in the kinds of output it produces, with available balance statistics include standardized mean differences, variance ratio, Kolmogorov-Smirnov statistics, and more. Utilities are available for computing both numeric and graphical displays of balance and make publication-ready plots for balance reporting. methods are available for assessing balance for binary, multi-category, continuous, and sequential treatments as well as clustered and multiply imputed data.

MatchThem: Matching and Weighting Multiply Imputed Datasets

Farhad Pishgar and Noah Greifer

CRAN Page | GitHub Page

MatchThem is a wrapper for functions in MatchIt and WeightIt for matching and weighting with multiply imputed data resulting from mice or Amelia. It is compatible with cobalt for balance assessment and provides utilities for applying Rubin’s rules for combining effect estimates after multiple imputation.

optweight: Targeted Stable Balancing Weights Using Optimization

Noah Greifer

CRAN Page | GitHub Page

optweight estimates balancing weights with minimal variance subject to user-specified constraints on covariate balance using convex optimization of quadratic programs. These weights can be used to achieve exact or approximate balance for binary, multi-category, and continuous treatments, as well as for matching adjusted indirect comparison of trials to target distributions. The methods implemented are described in Zubizarreta (2015), Wang and Zubizarreta (2020), and Greifer (2020).