Table 7:

Gaussian data with n = 200 and p = 51 under MNAR. Approximately 40% features and 92% cases contain missing values.

Models	Style	Time(s)	Imp MSE	Bias(${\widehat{\beta }}_1$)	CR(${\widehat{\beta }}_1$)	SE(${\widehat{\beta }}_1$)	SD(${\widehat{\beta }}_1$)
SoftImpute	SI	2.1	0.0119	−0.0053	0.93	0.0624	0.0611
GAIN	SI	35.9	1.4822	0.4448	0.24	0.1187	0.4641
MIWAE	SI	46.5	0.0361	−0.0238	0.90	0.0632	0.0738
not-MIWAE	SI	40.8	0.7566	−0.0436	0.91	0.0518	0.0969
Linear RR	SI	407	0.1760	0.0412	0.91	0.1314	0.1567
Sinkhorn	MI	27.9	0.1103	−0.1340	0.63	0.1006	0.1278
MICE	MI	2.1	0.0198	0.0036	0.98	0.0636	0.0559
MI-NNGP1	MI	4.7	0.0130	0.0027	0.95	0.0621	0.0651
MI-NNGP1-BS	MI	3.9	0.0177	0.0026	0.97	0.0799	0.0631
MI-NNGP2	MI	10.4	0.0088	0.0085	0.96	0.0614	0.0536
MI-NNGP2-BS	MI	10.1	0.0106	0.0093	0.97	0.0711	0.0564
Complete data	-	-	-	0.0025	0.98	0.0605	0.0524
Complete case	-	-	-	0.2143	0.78	0.2340	0.4201
ColMean Imp	SI	-	0.4772	0.5597	0.24	0.2246	0.1720