Table 4 Detailed results of model updating vs. full retraining
From: Bayesian parametric models for survival prediction in medical applications
Model | Experiment | C-Indexa,b | % in ROPE | HEquivalence |
|---|---|---|---|---|
DeepSurv | ACTG | 0.0610 [0.0520—0.0700] | 0.0 | Rejected |
BPS Exp | ACTG | 0.0026 [0.0007—0.0045] | 100.0 | Accepted |
BPS Wb | ACTG | 0.0023 [0.0006—0.0039] | 100.0 | Accepted |
DeepSurv | GBCS | -0.0004 [-0.0072—0.0062] | 100.0 | Accepted |
BPS Exp | GBCS | 0.0046 [0.0030—0.0063] | 100.0 | Accepted |
BPS Wb | GBCS | 0.0042 [0.0023—0.0060] | 100.0 | Accepted |
DeepSurv | PBC | 0.0260 [0.0190—0.0330] | 0.0 | Rejected |
BPS Exp | PBC | 0.0094 [0.0063—0.0130] | 68.1 | Undecided |
BPS Wb | PBC | 0.0088 [0.0054—0.0120] | 79.6 | Undecided |
DeepSurv | WHAS | 0.0290 [0.0140—0.0430] | 0.0 | Rejected |
BPS Exp | WHAS | 0.0052 [0.0046—0.0057] | 100.0 | Accepted |
BPS Wb | WHAS | 0.0035 [0.0030—0.0041] | 100.0 | Accepted |