Table 2:

Gaussian data with n = 200 and p = 251 under MAR. Approximately 40% features and 90% cases contain missing values. Detailed simulation setup information is in appendix.

Models	Style	Time(s)	Imp MSE	Bias(${\widehat{\beta }}_1$)	CR(${\widehat{\beta }}_1$)	SE(${\widehat{\beta }}_1$)	SD(${\widehat{\beta }}_1$)
SoftImpute	SI	15.1	0.0200	−0.0913	0.78	0.1195	0.1624
GAIN	SI	39.0	0.8685	0.6257	0.18	0.1463	0.5424
MIWAE	SI	46.3	0.0502	0.0731	0.90	0.1306	0.1379
Linear RR	SI	3134.7	0.0661	0.1486	1.00	0.1782	0.1011
MICE	MI	37.6	0.0234	−0.0061	0.93	0.1167	0.1213
Sinkhorn	MI	31.2	0.0757	0.0205	0.96	0.1864	0.1636
MI-NNGP1	MI	4.9	0.0116	0.0077	0.92	0.1147	0.1223
MI-NNGP1-BS	MI	3.4	0.0149	0.0156	0.96	0.1297	0.1182
MI-NNGP2	MI	5.7	0.0086	0.0012	0.96	0.1179	0.1170
MI-NNGP2-BS	MI	13.9	0.0094	0.0010	0.95	0.1173	0.1206
Complete data	-	-	-	−0.0027	0.90	0.1098	0.1141
Complete case	-	-	-	0.2481	0.88	0.3400	0.3309
ColMean Imp	SI	-	0.1414	0.3498	0.72	0.2212	0.1725