Table 1.

Results of treatment effect estimates from propensity matching and propensity weighting when assuming there is a homogeneous treatment effect and no unmeasured confounding. For each missing scenario, missing data are handled with complete case analysis, missing indicator method, multiple imputation, and the combination of multiple imputation and missing indicator (Combined method)

		Homogeneous treatment effect
		Propensity matching			Propensity weighting
		Coefficient		MSE	Coefficient		MSE
		Mean	SD		Mean	SD
No missing	No adjustment	1.298	0.123	1.700	1.298	0.123	1.700
	After adjustment	0.044	0.085	0.009	0.006	0.109	0.012
MCAR	Complete case analysis	0.043	0.121	0.016	0.014	0.152	0.023
	Missing indicator	0.238	0.095	0.066	0.189	0.111	0.048
	Multiple imputation
	With Y	0.047	0.086	0.010	0.011	0.113	0.013
	Without Y	0.219	0.087	0.056	0.186	0.110	0.047
	Combined method
	With Y	0.048	0.087	0.010	0.011	0.112	0.013
	Without Y	0.218	0.087	0.055	0.187	0.110	0.047
MAR	Complete case analysis	0.024	0.128	0.017	0.007	0.165	0.027
	Missing indicator	0.259	0.099	0.077	0.172	0.123	0.044
	Multiple imputation
	With Y	0.052	0.092	0.011	0.010	0.122	0.015
	Without Y	0.244	0.090	0.068	0.185	0.120	0.049
	Combined method
	With Y	0.050	0.092	0.011	0.010	0.122	0.015
	Without Y	0.243	0.090	0.067	0.185	0.120	0.048
MNAR	Complete case analysis	0.025	0.129	0.017	0.012	0.166	0.028
	Missing indicator	0.231	0.098	0.063	0.149	0.122	0.037
	Multiple imputation
	With Y	0.069	0.095	0.014	0.029	0.123	0.016
	Without Y	0.248	0.091	0.070	0.215	0.118	0.060
	Combined method
	With Y	0.052	0.093	0.011	0.011	0.122	0.015
	Without Y	0.211	0.088	0.053	0.160	0.119	0.040