Table 10:

Exponential data with n = 200 and p = 1001 under MNAR. Approximately 40% features and 92% cases contain missing values. Here Linear RR and MICE are not included due to running out of RAM.

Models	Style	Time(s)	Imp MSE	Bias(${\widehat{\beta }}_1$)	CR(${\widehat{\beta }}_1$)	SE(${\widehat{\beta }}_1$)	SD(${\widehat{\beta }}_1$)
SoftImpute	SI	39.1	0.6682	−0.6784	0.90	0.6805	0.4632
GAIN	SI	91.0	1.6974	0.0331	0.33	0.4187	2.2123
MIWAE	SI	57.4	1.4937	0.3981	0.30	0.1437	0.1659
not-MIWAE	SI	43.6	26.7277	0.7388	0.00	0.0928	0.1524
Sinkhorn	MI	53.2	0.3837	−0.2698	1.0	0.4362	0.2886
MI-NNGP1	MI	3.5	0.3296	−0.0354	0.76	0.1764	0.3166
MI-NNGP1-BS	MI	3.6	0.4545	−0.0546	0.99	0.3324	0.2466
MI-NNGP2	MI	4.1	0.2360	0.0835	0.82	0.2301	0.3327
MI-NNGP2-BS	MI	10.3	0.2501	−0.0449	0.94	0.3183	0.3043
Complete data	-	-	-	0.0156	0.95	0.0978	0.0938
Complete case	-	-	-	0.1663	0.89	0.4038	0.4639
ColMean Imp	SI	-	0.3793	−0.1574	1.0	0.4559	0.3087