Skip to main content

Table 3 Median (minimum, maximum) number of correctly identified non-zero and zero coefficients, N = 1000, 1000 simulated datasets

From: Least absolute shrinkage and selection operator type methods for the identification of serum biomarkers of overweight and obesity: simulation and application

Scenario 1 2 3 4 5
Type Non-Zero Zero Non-Zero Zero Non-Zero Zero Non-Zero Zero Non-Zero Zero
Truth 5 95 5 95 5 95 5 95 20 80
Overweight
 LASSO 3 (0,5) 91 (61,95) 3 (1,5) 92 (56,95) 5 (4,5) 81 (39,95) 4 (2,5) 89 (43,95) 13 (8,18) 74.5 (39,80)
 Adaptive LASSO 3 (0,5) 82 (54,95) 2 (1,5) 83 (50,95) 5 (4,5) 82 (51,95) 4 (2,5) 81 (48,95) 9 (4,15) 68 (41,80)
 Elastic Net 3 (0,5) 86 (0,95) 3 (1,5) 88 (0,95) 5 (4,5) 77 (38,95) 4 (2,5) 86 (34,95) 16 (9,20) 67 (0,80)
 Iterated LASSO 2 (0,5) 91 (57,95) 2 (1,5) 93 (64,95) 5 (4,5) 82 (54,95) 3 (2,5) 90 (57,95) 11 (5,16) 76 (50,80)
 Bootstrap-Enhanced LASSO-75 3 (0,5) 86 (70,95) 2 (0,4) 86 (72,95) 5 (4,5) 82 (62,93) 4 (2,5) 81 (60,94) 12 (7,16) 72 (54,80)
 Weighted Fusion 4 (0,5) 72 (1,95) 5 (1,5) 72 (0,95) 5 (4,5) 77 (24,90) 4 (2,5) 77 (26,89) 20 (9,20) 65 (0,74)
Obese
 LASSO 5 (3,5) 82 (49,95) 3 (2,5) 91 (41,95) 5 (5,5) 74 (44,92) 4 (3,5) 83 (42,95) 19 (14,20) 61 (37,78)
 Adaptive LASSO 5 (3,5) 82.5 (50,95) 4 (1,5) 82 (53,95) 5 (5,5) 82 (43,95) 5 (3,5) 81 (53,95) 15 (10,20) 60 (38,79)
 Elastic Net 5 (4,5) 79 (26,94) 4 (2,5) 88 (25,95) 5 (5,5) 71 (24,94) 4 (3,5) 80 (23,95) 19 (15,20) 52 (0,75)
 Iterated LASSO 5 (3,5) 84 (60,95) 3 (1,5) 91 (56,95) 5 (5,5) 80 (60,95) 4 (3,5) 86 (62,95) 16 (12,20) 72 (56,80)
 Bootstrap-Enhanced LASSO-75 5 (3,5) 83 (63,95) 3 (1,5) 84 (64,95) 5 (5,5) 80 (61,93) 4 (2,5) 83 (66,94) 16 (10,20) 73 (62,80)
 Weighted Fusion 5 (3,5) 88 (32,94) 3 (2,5) 89 (1,95) 5 (5,5) 57 (39,95) 4 (3,5) 64 (46,95) 20 (14,20) 61 (1,76)