Random Recursive Partitioning: a matching method for the estimation of the average treatment effect

摘要: In this paper we introduce the Random Recursive Partitioning (RRP) matching method. RRP generates a proximity matrix which might be useful in econometric applications like average treatment effect estimation. is Monte Carlo method that randomly non-empty recursive partitions of data and evaluates between two observations as empirical frequency they fall same cell these random over all replications. From it possible to derive both graphical analytical tools evaluate extent common support sets. The “honest” does not match “at any cost”: if sets are separated, clearly states it. The obtained with invariant under monotonic transformation data. Average estimators derived from seem competitive compared more commonly used estimators. require particular structure for reason can applied when distances Mahalanobis or Euclidean suitable, presence missing estimated propensity score too sensitive model specifications. Copyright © 2008 John Wiley & Sons, Ltd.