A caliper is a maximum distance that two units can be apart from each other (on their estimated propensity scores) and is defined in units of standard deviations of the logit of the estimated propensity score. Defining a small caliper will usually result in better balance at the expense of finding fewer units that can be successfully matched. Conversely, a large caliper will retain more matches, but some of them will be slightly imbalanced, and might yield a larger bias in the estimation of the treatment effect.For unweighted data, an overall imbalance χ2 test is provided。After matching is completed, a series of model adequacy checks should be performed.The main interest of the researcher is to check whether balance on the covariates has truly been achieved through the matching procedure. This can be done by comparing several statistics of the treatment and control group before and after matching, most often the standardized mean differences and the variance ratio. The standardized mean difference of covariates should be close to 0 after matching, and the variance ratio should be close to 1. In addition, bootstrapped Kolomgorov-Smirnov tests can be computed to examine equality of distribution of single covariatesand global imbalance measures or multivariate significance tests can be computed and assessed.more psm
Propensity score matching in SPSS.pdf
(719.37 KB)
发帖
雷达卡
jg-xs1
京公网安备 11010802022788号
论坛法律顾问：王进律师
知识产权保护声明
免责及隐私声明
