Research Article
Comparison of Variable Selection Methods for Time-to-Event Data in High-Dimensional Settings
Table 2
Number of selected predictors for each simulated dataset and selection method.
| | | Training fraction | Number of selected predictors
med (min–max) | | LASSO-cvl | LASSO-pcvl | Elastic Net | BSS Enet | CoxBoost | RSF |
| | 500 | 0 | 1/2 | 0 (0-24) | 1 (0-30) | 0 (0-34) | 9 (1-15) | 0 (0-24) | 71 (67-75) | | 2/3 | 0 (0-21) | 0 (0-31) | 0 (0-24) | 10 (4-20) | 0 (0-24) | 56 (52-59) | | 12 | 1/2 | 25 (7-55) | 11 (5-29) | 33 (10-62) | 22 (10-29) | 26 (9-50) | 71 (65-75) | | 2/3 | 27 (10-55) | 13 (7-26) | 37 (13-75) | 27 (19-35) | 30 (10-50) | 56 (52-60) | | 50 | 1/2 | 62 (40-93) | 44 (13-75) | 83 (50-121) | 44 (29-57) | 57 (24-80) | 75 (70-78) | | 2/3 | 76 (57-105) | 53 (34-72) | 96 (71-127) | 56 (47-66) | 67 (41-103) | 63 (57-68) |
| | 750 | 0 | 1/2 | 0 (0-18) | 0 (0-33) | 0 (0-23) | 12 (6-22) | 0 (0-29) | 45 (42-49) | | 2/3 | 0 (0-20) | 0 (0-16) | 0 (0-21) | 16 (8-26) | 0 (0-19) | 22 (19-25) | | 12 | 1/2 | 25 (12-52) | 13 (7-25) | 33 (14-57) | 25 (14-36) | 27 (12-53) | 47 (44-51) | | 2/3 | 24 (12-49) | 11 (8-18) | 28 (17-55) | 28 (21-37) | 26 (13-52) | 25 (22-28) | | 50 | 1/2 | 72 (51-93) | 44 (27-68) | 88 (65-118) | 54 (46-69) | 62 (36-85) | 58 (53-61) | | 2/3 | 80 (58-111) | 46 (31-59) | 96 (76-134) | 64 (52-78) | 69 (43-89) | 37 (32-41) |
| | 1000 | 0 | 1/2 | 0 (0-45) | 0 (0-36) | 0 (0-45) | 19 (10-34) | 1 (0-40) | 22 (20-25) | | 2/3 | 1 (0-48) | 1 (0-43) | 1 (0-48) | 27 (17-40) | 2 (0-45) | 6 (6-8) | | 12 | 1/2 | 34 (11-79) | 15 (8-28) | 42 (18-89) | 35 (24-52) | 36 (17-62) | 28 (25-32) | | 2/3 | 42 (21-74) | 15 (9-25) | 50 (23-86) | 48 (31-63) | 44 (14-77) | 9 (8-11) | | 50 | 1/2 | 90 (70-124) | 57 (34-78) | 107 (87-139) | 68 (58-89) | 74 (54-114) | 38 (34-43) | | 2/3 | 101 (76-145) | 57 (46-73) | 120 (84-168) | 80 (67-99) | 79 (53-112) | 18 (16-20) |
|
|
