Table 1. Simulation comparing empirical size of the goodness-of-fit test obtained using our model-based approach and the extrapolation method with 6 risk score groups and α=0.05 for an exponential baseline hazard function.
n | Censoring % | Type 1 error: Model-based approach | Type 1 error: Extrapolation approach | Mean observed events | Mean projected events in extrapolation approach |
---|---|---|---|---|---|
500 | 10% | 0.031 | 0.054 | 314 | 338 |
500 | 20% | 0.035 | 0.080 | 301 | 338 |
500 | 30% | 0.018 | 0.161 | 272 | 338 |
500 | 40% | 0.028 | 0.276 | 256 | 338 |
500 | 50% | 0.030 | 0.591 | 232 | 338 |
| |||||
200 | 10% | 0.025 | 0.063 | 124 | 134 |
200 | 20% | 0.026 | 0.088 | 119 | 134 |
200 | 30% | 0.026 | 0.213 | 108 | 134 |
200 | 40% | 0.038 | 0.329 | 101 | 134 |
200 | 50% | 0.031 | 0.573 | 92 | 134 |
results based on 1000 replications for each simulation