This is a critical distinction because matching imputation is a specific estimation method with its own effect and standard error estimators, in contrast to subset selection, which is a preprocessing method that does not require specific estimators and is broadly compatible with other parametric and nonparametric analyses. That form of matching is matching imputation, where the missing potential outcomes for each unit are imputed using the observed outcomes of paired units.
It is important to note that this implementation of matching differs from the methods described by Abadie and Imbens ( 2006, 2016) and implemented in the Matching R package and teffects routine in Stata. These advantages, as well as the intuitive understanding of matching by the public compared to regression or weighting, make it a robust and effective way to estimate treatment effects. Matching is nonparametric in the sense that the estimated weights and pruning of the sample are not direct functions of estimated model parameters but rather depend on the organization of discrete units in the sample this is in contrast to propensity score weighting (also known as inverse probability weighting), where the weights come more directly from the estimated propensity score model and therefore are more sensitive to its correct specification. ( 2007) argue that fitting regression models in matched samples reduces the dependence of the validity of the estimated treatment effect on the correct specification of the model. Although statistical estimation methods like regression can also be used to remove confounding due to measured covariates, Ho et al. Ideally, and if done successfully, subset selection produces a new sample where the treatment is unassociated with the covariates so that a comparison of the outcomes treatment and control groups is not confounded by the measured and balanced covariates. Matching as implemented in MatchIt is a form of subset selection, that is, the pruning and weighting of units to arrive at a (weighted) subset of the units from the original dataset.
#Propensity score matching spss version 25 how to#
For details on how to estimate treatment effects and standard errors after matching, see vignette("estimating-effects"). For details on how to assess and report covariate balance, see vignette("assessing-balance"). For a brief introduction to the use of MatchIt functions, see vignette("MatchIt").
This vignette describes each matching method available in MatchIt and the various options that are allowed with matching methods and the consequences of their use. A benefit of nonparametric preprocessing through matching is that a number of matching methods can be tried and their quality assessed without consulting the outcome, reducing the possibility of capitalizing on chance while allowing for the benefits of an exploratory analysis in the design phase ( Ho et al. The choice of matching method depends on the goals of the analysis (e.g., the estimand, whether low bias or high precision is important) and the unique qualities of each dataset to be analyzed, so there is no single optimal choice for any given analysis. Though the help pages for the individual methods describes each method and how they can be used, this vignette provides a broad overview of the available matching methods and their associated options.
MatchIt implements several matching methods with a variety of options.