Table 2.

Simulation results for MAR scenarios with different missing patterns (independent and monotone) and different correlation structures (compound symmetry with low correlation, compound symmetry with medium correlation and first-order autoregressive (AR(1))) for p = 12 and n = 50.

	Independent MAR				Monotone MAR
	PMSE	MSE	SENS	SPEC	PMSE	MSE	SENS	SPEC
	Low correlation: corr(Xi,j ,Xi,j′ ) = 0.1
Full-SPLS	14.45	3.33	94.8	75.7	14.66	3.54	95.3	70.5
CC-SPLS	27.27	12.69	78.7	59.2	33.38	14.95	72.5	56.3
SI-SPLS	18.03	6.61	89.3	68.0	19.66	8.19	84.0	66.0
MI-SPLS	16.52	4.93	98.0	38.7	17.18	5.72	95.0	44.3
Full-ENet	13.83	2.71	96.8	78.8	13.93	2.88	97.2	75.2
CC-ENet	26.83	12.82	64.5	92.0	31.14	15.20	58.8	91.5
SI-ENet	16.90	5.73	90.7	77.2	18.25	7.07	85.2	72.8
MI-ENet	16.38	5.07	92.8	74.7	16.88	5.70	90.0	75.5
SI-WENet	16.57	5.35	91.3	78.3	17.20	5.96	88.7	73.5
MI-WENet	15.96	4.74	93.5	76.5	16.41	5.23	89.8	73.7
	Medium correlation: corr(Xi,j ,Xi,j′ ) = 0.5
Full-SPLS	32.28	7.24	90.5	34.5	32.77	7.67	89.0	38.0
CC-SPLS	40.92	16.02	81.3	40.2	49.39	19.42	79.8	38.2
SI-SPLS	35.15	9.57	80.7	45.0	35.49	9.47	81.3	41.5
MI-SPLS	34.69	8.83	91.5	30.8	34.88	8.69	90.7	34.3
Full-ENet	31.27	6.27	83.3	72.8	30.76	6.01	84.2	73.2
CC-ENet	53.89	27.42	47.5	91.3	58.18	28.58	43.8	88.8
SI-ENet	33.21	8.31	77.3	70.5	33.31	8.17	76.2	69.3
MI-ENet	32.72	7.79	78.2	69.0	32.22	7.20	77.5	68.2
SI-WENet	33.47	8.57	76.8	71.3	33.00	8.06	75.8	71.7
MI-WENet	32.63	7.66	78.8	69.0	31.94	6.99	78.2	67.8
	AR(1) correlation: corr(Xi,j ,Xi,j′ ) = 0.8|j-j′ |
Full-SPLS	27.36	5.66	89.7	30.2	27.45	5.82	87.5	31.0
CC-SPLS	33.11	11.68	81.0	36.3	43.08	15.44	75.8	35.8
SI-SPLS	30.46	8.28	81.3	40.3	30.33	7.31	81.3	42.0
MI-SPLS	30.96	8.46	91.0	23.5	30.02	7.15	89.2	27.3
Full-ENet	26.91	5.19	78.5	66.8	26.59	5.04	78.5	70.5
CC-ENet	45.50	22.77	41.8	84.5	53.26	28.26	38.0	87.2
SI-ENet	30.21	8.30	67.8	70.5	28.28	6.68	69.3	72.2
MI-ENet	28.76	6.93	72.8	69.2	27.76	6.05	72.7	70.5
SI-WENet	28.81	7.01	70.3	67.5	27.95	6.37	69.0	72.0
MI-WENet	27.97	6.34	73.8	68.8	27.22	5.63	73.5	69.8