To provide an alternative to propensity scoring () for the common situation where there are interacting covariates.
We used 1.3 million assessments of residents of the United States Veterans Affairs nursing homes, collected from January 1, 2000, through October 9, 2012.
In stratified covariate balancing (), data are divided into naturally occurring strata, where each stratum is an observed combination of the covariates. Within each stratum, cases with, and controls without, the target event are counted; controls are weighted to be as frequent as cases. This weighting procedure guarantees that covariates, or combination of covariates, are balanced, meaning they occur at the same rate among cases and controls. Finally, impact of the target event is calculated in the weighted data. We compare the performance of , logistic regression (), and propensity scoring () in simulated and real data. We examined the calibration of and in predicting 6‐month mortality from inability to eat, controlling for age, gender, and nine other disabilities for 296,051 residents in Veterans Affairs nursing homes. We also performed a simulation study, where outcomes were randomly generated from treatment, 10 covariates, and increasing number of covariate interactions. The accuracy of ,, and in recovering the simulated treatment effect was reported.
In simulated environment, as the number of interactions among the covariates increased, and properly specified remained accurate but pairwise and pairwise , the most common applications of these tools, performed poorly. In real data, application of was practical. was better calibrated than linear , the most common method of .
In environments where covariates interact, is practical and more accurate than common methods of applying and .