Categorical covariates, scales well to large datasets with millions of observations
Matching method:
Uses bit-vector computations to match units based on a learned, weighted Hamming distance. FLAME successively drops irrelevant covariates to lessen the computational load while still maintaining enough covariates for high-quality conditional average treatment effect (CATE) estimation.
Continuous, categorical, or mixed (continous and categorical) covariates
Matching method:
Uses exact matching for discrete variables and learned, generalized Mahalanobis distances for continuous variables. Instead of a predetermined distance metric, the covariates contributing more towards predicting the outcome are given higher weights.
Continuous, categorical, or mixed (continous and categorical) covariates
Matching method:
Matches units with others in unit-specific, hyper-box-shaped regions of the covariate space. The regions are found as either the solution to a mixed integer program, or by using a fast approximation algorithm.
Creates almost exact matches in a computationally scalable manner, which works well for high-dimensional and/or big data. The feature importances from an outcome regression model are used as distance metric weights.