Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure

Coffman, Donna L.; Zhou, Jiangxiu; Cai, Xizhen

doi:10.1186/s12874-020-01053-4

BMC Medical Research Methodology

Table 7 Simulation results for scenario G with the correct logistic model (25%missing)

From: Comparison of methods for handling covariate missingness in propensity score estimation with a binary exposure

Missingness Mechanism	Method	n = 500			n = 1000			n = 5000
Missingness Mechanism	Method	Bias (SD)	SE	RMSE	Bias (SD)	SE	RMSE	Bias (SD)	SE	RMSE
	complete	−0.002 (0.045)	0.070	0.045	0.000 (0.030)	0.049	0.030	0.001 (0.014)	0.022	0.014
MCAR	SI + pe + pu	−0.002 (0.053)	0.070	0.053	−0.001 (0.035)	0.049	0.035	0.000 (0.016)	0.022	0.016
	SI + pe	−0.003 (0.052)	0.070	0.052	0.000 (0.035)	0.049	0.035	0.000 (0.016)	0.022	0.016
	TMI	−0.013 (0.062)	0.076	0.063	−0.009 (0.041)	0.053	0.042	−0.008 (0.018)	0.023	0.020
	MI	−0.002 (0.048)	0.071	0.048	−0.001 (0.033)	0.049	0.033	0.000 (0.015)	0.022	0.014
	MIMP	−0.002 (0.048)	0.071	0.049	−0.001 (0.033)	0.050	0.033	0.000 (0.015)	0.022	0.014
MAR1	SI + pe + pu	0.009 (0.052)	0.070	0.053	0.009 (0.035)	0.049	0.036	0.010 (0.016)	0.022	0.019
	SI + pe	0.008 (0.051)	0.070	0.052	0.010 (0.034)	0.049	0.035	0.011 (0.016)	0.022	0.019
	TMI	0.019 (0.057)	0.073	0.060	0.021 (0.039)	0.051	0.044	0.023 (0.018)	0.023	0.029
	MI	0.008 (0.047)	0.070	0.048	0.009 (0.031)	0.049	0.033	0.010 (0.015)	0.022	0.017
	MIMP	0.008 (0.048)	0.071	0.048	0.009 (0.032)	0.049	0.034	0.009 (0.015)	0.022	0.017
MAR2	SI + pe + pu	−0.001 (0.052)	0.070	0.052	0.003 (0.036)	0.049	0.036	0.002 (0.016)	0.022	0.016
	SI + pe	0.000 (0.052)	0.070	0.052	0.003 (0.035)	0.049	0.035	0.002 (0.016)	0.022	0.016
	TMI	−0.019 (0.064)	0.077	0.067	−0.015 (0.043)	0.054	0.046	−0.015 (0.019)	0.024	0.024
	MI	0.000 (0.047)	0.071	0.048	0.003 (0.032)	0.049	0.033	0.002 (0.015)	0.022	0.014
	MIMP	0.001 (0.047)	0.071	0.046	0.004 (0.033)	0.050	0.033	0.005 (0.015)	0.022	0.015
MAR sinister	SI + pe + pu	0.003 (0.050)	0.070	0.050	0.002 (0.035)	0.049	0.035	0.000 (0.016)	0.022	0.016
	SI + pe	−0.003 (0.052)	0.070	0.052	0.002 (0.035)	0.049	0.035	0.000 (0.016)	0.022	0.016
	TMI	0.002 (0.061)	0.076	0.061	0.003 (0.042)	0.052	0.042	−0.004 (0.018)	0.023	0.018
	MI	0.001 (0.047)	0.071	0.048	0.002 (0.033)	0.049	0.032	0.000 (0.015)	0.022	0.015
	MIMP	0.001 (0.047)	0.071	0.047	0.002 (0.033)	0.049	0.033	0.000 (0.015)	0.022	0.014

Note. Complete: logistic regression with complete data before introducing missingness; SI + pe + pu single imputation + prediction error + parameter uncertainty; SI + pe single imputation + prediction error; TMI treatment mean imputation; MI multiple imputation (m = 20); MIMP multiple imputation missingness pattern (m = 20); SD standard deviation; SE standard error; RMSE root mean squared error

Back to article page

ISSN: 1471-2288

Contact us

General enquiries: journalsubmissions@springernature.com