Is what justifies ignorability on the observed covariates, allowing for Achievingīalance is the very purpose of preprocessing because covariate balance Propensity score applications, with researchers failing to report theĭegree of covariate balance achieved by their conditioning ( Thoemmes and Kim 2011). However, Step 4, assessing balance, is often overlooked in Steps 2, 3, and 4 are accomplished by all of the packages mentionedĪbove. Estimate the treatment effect in the conditioned sample.Assess balance on the covariates of interest if poor, repeat steps.Condition on the distance measure (e.g., using matching, weighting,.Estimate the distance measure (e.g., propensity score).

Decide on covariates for which balance must be achieved.

The following are the basic steps in performing a causal analysis King, and Porro 2009) these together provide a near complete 2019), ebal ( Hainmueller 2014), sbw ( Zubizarreta, Li, and Kim 2021),ĭesignmatch ( Zubizarreta, Kilcioglu, and VielmaĢ018), WeightIt ( Greifer 2021), MatchThem ( Pishgar et al. Matching ( Sekhon 2011), optmatch ( Hansen and Klopfer 2006),Īl. Several packages in R exist to perform preprocessing and causalĮffect estimation, and some were reviewed by Keller and Tipton ( 2016). Of the causal assumption of strong ignorability is assumed. The logic and theoryīehind preprocessing will not be discussed here, and reader’s knowledge

Introductions to propensity scores and other preprocessing methods areĬlearly and point to other sources of knowledge. Other related methods (e.g., coarsened exact matching, Mahalanobisĭistance matching, genetic matching) have become popular in the socialĪnd health sciences as tools for this purpose. Preprocessing data through matching, weighting, or subclassificationĬan be an effective way to reduce model dependence and improveĮfficiency when estimating the causal effect of a treatment ( Ho et al.

Standard entropy