A Fast Large-Scale Almost Matching Exactly Approach to Causal Inference

The 'FLAME' (Fast Large-scale Almost Matching Exactly) package implements the matching algorithm in Roy, Rudin, Volfovsky, and Wang (2017) . 'FLAME' performs matching of treatment and control units in the potential outcomes framework for large categorical datasets. 'FLAME' creates matches that include as many covariates as possible, and sequentially drops covariates that are less useful based on a match quality measure. Match quality combines two important elements – it considers predictive power from machine learning on a hold out training set, and a balancing factor to ensure that it does not remove a covariate that would ruin overlap between treatment and control groups. Currently the 'FLAME' package applies to categorical data, and provides two approaches for implementation - bit vectors and database management systems (e.g., 'PostgreSQL', 'SQLite'). For data that has been adequately processed and fits in memory, bit vectors should be applied. For extremely large data that does not fit in memory, database systems should be applied.


Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


1.0.0 by Chia-Rui (Jerry) Chang, 2 months ago


Report a bug at https://github.com/chiarui424/FLAME/issues

Browse source code at https://github.com/cran/FLAME

Authors: Chia-Rui (Jerry) Chang [aut, cre] , Cynthia Rudin [aut] , Alexander Volfovsky [aut] , Sudeepa Roy [aut] , Tianyu Wang [ctb]

Documentation:   PDF Manual  

GPL-3 license

Imports reticulate, graphics, stats, latticeExtra, dplyr, RPostgreSQL, RSQLite, gmp, lattice, rlang

Suggests knitr, rmarkdown, ggplot2, ggpubr, e1071

See at CRAN