Table 2.
Prediction accuracy at landmark times of the proposed nonparametric kernel Nelson-Aalen estimator (NPNA) compared to the estimator that ignores landmarking (NPNA estimator assuming
, denoted as NPNA
), and the Kaplan–Meier estimator that ignores covariate influence (KM) and modified to adjust for landmarking
| KM | NPNA | NPNA
|
NPNA-KM
|
NPNA-NPNA
|
|
|---|---|---|---|---|---|
AUC, Setting 1:
|
|||||
|
51.2854 | 64.4163 | 61.4384 | (7.4368, 18.8251) | ( 0.6716, 6.6274) |
|
55.1505 | 64.5175 | 62.8209 | (5.0543, 13.6799) | ( 0.3956, 3.7888) |
|
57.652 | 65.9349 | 64.7988 | (4.0615, 12.5044) | ( 0.5791, 2.8514) |
BS, Setting 1:
|
|||||
|
13.5879 | 13.084 | 15.2423 | ( 0.9247, 0.0831) |
( 2.8356, 1.4809) |
|
23.3485 | 22.0813 | 23.2354 | ( 2.0813, 0.4531) |
( 1.7654, 0.5428) |
|
24.369 | 22.9809 | 23.6991 | ( 2.3385, 0.4377) |
( 1.2936, 0.1428) |
AUC, Setting 1:
|
|||||
|
56.718 | 64.7069 | 61.1501 | (3.0307, 12.9472) | (0.204, 6.9096) |
|
58.876 | 65.9604 | 63.6971 | (2.7056, 11.4633) | ( 0.2317, 4.7584) |
BS, Setting 1:
|
|||||
|
19.0458 | 18.2571 | 22.9205 | ( 1.5142, 0.0632) |
( 5.7315, 3.5954) |
|
23.6235 | 22.4671 | 25.2915 | ( 2.1805, 0.1323) |
( 3.833, 1.816) |
AUC, Setting 2:
|
|||||
|
68.672 | 82.9663 | 83.0024 | (9.9418, 18.6468) | ( 1.3709, 1.2988) |
|
69.2237 | 82.9825 | 83.0218 | (10.1455, 17.372) | ( 0.5834, 0.5048) |
|
70.476 | 83.3167 | 83.3574 | (8.9964, 16.685) | ( 0.4441, 0.3628) |
BS, Setting 2:
|
|||||
|
15.7572 | 12.7593 | 12.8608 | ( 4.0232, 1.9726) |
( 0.5754, 0.3724) |
|
21.0976 | 16.6319 | 16.5867 | ( 5.8972, 3.0343) |
( 0.2675, 0.3579) |
|
18.859 | 15.3791 | 15.3327 | ( 4.7729, 2.1868) |
( 0.1358, 0.2287) |
AUC, Setting 2:
|
|||||
|
66.1443 | 78.9501 | 79.3607 | (6.4, 19.2115) | ( 4.5741, 3.7529) |
|
66.8676 | 79.7724 | 79.9264 | (9.0036, 16.806) | ( 1.3163, 1.0083) |
|
68.1474 | 80.6619 | 80.7738 | (8.7062, 16.3229) | |
BS, Setting 2:
|
|||||
|
9.2491 | 8.3604 | 10.3928 | ( 1.5438, 0.2337) |
( 3.2139, 0.8508) |
|
20.6777 | 17.2407 | 17.3808 | ( 4.8153, 2.0585) |
( 0.9188, 0.6387) |
|
20.6045 | 17.2658 | 17.1442 | ( 4.6848, 1.9927) |
( 0.299, 0.5421) |
These columns correspond to the 95% bootstrap confidence intervals for the differences between estimators. We report the AUC and the BS for
at specified
and landmark times
. Results are summarized over 200 simulations with 40% censoring. Note that better prediction accuracy is indicated by higher AUC and lower BS. Sample size is 2000. All results are multiplied by 100.


























































