Outcome-sensitive multiple imputation: a simulation study

Kontopantelis, Evangelos; White, Ian R.; Sperrin, Matthew; Buchan, Iain

doi:10.1186/s12874-016-0281-5

BMC Medical Research Methodology

Table 2 Performance results for exposure E, datasets of 1,000 observations^a

From: Outcome-sensitive multiple imputation: a simulation study

		% miss	A	B	C^b	D^b	E^b	F^b	G^b	H
MCAR	Mean bias^c	20	−0.019	−0.107	−0.020	−0.025	−0.030	−0.021	−0.023	−0.427
		40	−0.038	−0.209	−0.041	−0.045	−0.020	−0.041	−0.045	−0.437
		60	−0.084	−0.301	−0.064	−0.056	−0.060	−0.067	−0.055	−0.440
		80	0.134	−0.409	−0.109	−0.092	−0.145	−0.126	−0.097	−0.458
	Mean error^c	20	0.218	0.208	0.200	0.201	0.199	0.199	0.199	0.430
		40	0.290	0.268	0.232	0.231	0.229	0.232	0.233	0.444
		60	0.475	0.361	0.306	0.313	0.308	0.309	0.301	0.466
		80	1.016	0.503	0.584	0.503	0.543	0.560	0.489	0.525
	Coverage	20	0.950	0.941	0.943	0.945	0.949	0.947	0.943	0.489
		40	0.962	0.913	0.963	0.957	0.959	0.962	0.949	0.605
		60	0.965	0.907	0.959	0.939	0.955	0.960	0.943	0.718
		80	0.992	0.957	0.989	0.956	0.986	0.988	0.965	0.823
	Power	20	0.720	0.688	0.771	0.761	0.777	0.769	0.770	0.274
		40	0.462	0.425	0.604	0.593	0.644	0.610	0.597	0.181
		60	0.228	0.213	0.367	0.413	0.394	0.373	0.433	0.163
		80	0.065	0.062	0.068	0.180	0.098	0.074	0.169	0.124
MAR	Mean bias^c	20	−0.019	−0.125	−0.030	−0.035	−0.030	−0.030	−0.035	−0.425
		40	−0.014	−0.227	−0.046	−0.065	−0.044	−0.050	−0.062	−0.441
		60	−0.046	−0.308	−0.063	−0.064	−0.049	−0.059	−0.062	−0.446
		80	−0.054	−0.350	−0.069	−0.119	−0.080	−0.073	−0.128	−0.408
	Mean error^c	20	0.224	0.215	0.200	0.200	0.203	0.198	0.197	0.428
		40	0.290	0.282	0.228	0.227	0.234	0.229	0.229	0.448
		60	0.491	0.367	0.314	0.312	0.303	0.314	0.305	0.467
		80	1.070	0.463	0.593	0.501	0.546	0.601	0.502	0.494
	Coverage	20	0.941	0.925	0.955	0.954	0.956	0.955	0.953	0.504
		40	0.958	0.896	0.953	0.956	0.950	0.958	0.948	0.561
		60	0.955	0.889	0.963	0.948	0.971	0.964	0.946	0.706
		80	0.994	0.943	0.978	0.954	0.978	0.978	0.946	0.836
	Power	20	0.708	0.678	0.779	0.763	0.771	0.774	0.779	0.271
		40	0.448	0.372	0.583	0.564	0.606	0.587	0.593	0.193
		60	0.224	0.195	0.350	0.391	0.373	0.360	0.404	0.161
		80	0.010	0.052	0.050	0.147	0.077	0.045	0.143	0.123
MNAR	Mean bias^c	20	−0.026	−0.150	−0.047	−0.044	−0.054	−0.047	−0.046	−0.425
		40	−0.092	−0.302	−0.135	−0.113	−0.097	−0.139	−0.121	−0.454
		60	−0.086	−0.334	−0.050	−0.022	−0.027	−0.052	−0.021	−0.450
		80	0.038	−0.478	−0.253	−0.283	−0.314	−0.316	−0.275	−0.484
	Mean error^c	20	0.227	0.228	0.207	0.207	0.209	0.208	0.207	0.431
		40	0.375	0.371	0.302	0.293	0.292	0.304	0.293	0.472
		60	0.590	0.411	0.366	0.368	0.361	0.371	0.355	0.479
		80	1.283	0.741	0.841	0.737	0.773	0.881	0.711	0.654
	Coverage	20	0.954	0.923	0.940	0.945	0.941	0.944	0.943	0.554
		40	0.957	0.927	0.958	0.950	0.961	0.954	0.949	0.708
		60	0.973	0.962	0.967	0.951	0.964	0.968	0.946	0.746
		80	0.997	1.000	0.993	0.990	0.999	0.995	0.986	0.921
	Power	20	0.652	0.592	0.714	0.712	0.694	0.706	0.713	0.237
		40	0.307	0.223	0.383	0.398	0.420	0.389	0.399	0.153
		60	0.209	0.148	0.330	0.368	0.366	0.327	0.382	0.152
		80	0.015	0.010	0.041	0.088	0.039	0.030	0.096	0.092

^aAnalysis model A: complete case analysis (no multiple imputation [mi]); B: no outcome imputation, not included in mi model; C: no outcome imputation, outcome imputed in mi model; D: outcome imputed and included in mi model; E: outcome imputed and included in mi model but then observations where it was imputed are deleted; F as in C but also including a second correlated outcome in the mi model; G as in D but also including a second correlated outcome in the mi model; H as in D but the mi and analysis models do not include the covariate
^bMain models of interest, other models provided for comparison purposes
^cReported on log-odds scale and based on a true effect of log [2]

Back to article page

ISSN: 1471-2288

Contact us

General enquiries: journalsubmissions@springernature.com