Table 3.

Biases (“bias”), ratios of empirical SEs (“ratio SE”), ratios of mean estimated SEs (“ratio empSE”), and relative efficiencies (%, “rel. eff.”) of six MI methods when N = 500 and p_miss = 0.25. Ratios and relative efficiencies are calculated relative to the corresponding complete-data estimators. Each reported ratio or relative efficiency is the average over three ratios or relative efficiencies: one from each of the MCAR, MAR-A, and MAR-B scenarios. Reported biases are the signed average absolute bias over these three scenarios.

	βcat				βconA
	bias	ratio SE	ratio estSE	rel. eff.	bias	ratio SE	ratio estSE	rel. eff.
M=1
Complete data	0.010	1.000	1.000	100.0	0.003	1.000	1.000	100.0
Match var: FCS	0.014	1.136	1.188	77.4	0.002	1.142	1.161	76.7
Normal	−0.046	1.018	1.156	92.2	0.006	1.137	1.157	77.2
Latent norm	0.026	1.126	1.183	78.1	−0.010	1.081	1.150	84.4
Match set: FCS	0.013	1.181	1.207	71.7	−0.002	1.196	1.184	70.0
Normal	−0.015	1.132	1.245	77.9	0.014	1.251	1.223	62.7
Latent norm	0.035	1.187	1.220	69.7	−0.035	1.082	1.157	73.0
M=4
Complete data	0.001	1.000	1.000	100.0	0.001	1.000	1.000	100.0
Match var: FCS	−0.001	1.195	1.200	70.0	0.002	1.188	1.172	70.8
Normal	−0.051	1.108	1.171	73.9	0.005	1.184	1.169	70.9
Latent norm	0.010	1.177	1.195	72.0	−0.010	1.129	1.163	76.6
Matchset: FCS	−0.006	1.220	1.210	67.3	−0.002	1.190	1.187	70.5
Normal	−0.036	1.149	1.206	72.0	0.007	1.215	1.196	67.0
Latent norm	0.008	1.200	1.211	69.3	−0.034	1.126	1.168	62.4