Propensity Scoring with Missing Values
Propensity scores have been used widely as a bias reduction method to estimate the treatment effect in nonrandomized studies. Because many covariates are generally included in the model for estimating propensity scores, the proportion of subjects with at least one missing covariate can be relatively large. In this chapter, we review existing methods for estimating propensity scores when missing values are present. The methods include a complete covariate (CC) method, an indicator variable (IND) method, various multiple ...