56
Table 3.1: Results Obtained Using the Proposed Method With τ=0.8; T, C and S
are 47%, 48.5% and 4.5%.
Parameters |
True Values |
Estimates |
SD of Est |
SE of Est | |
1 |
βι |
.5 |
0.543 |
0.169 |
0.204 |
2 |
βl |
.4 |
0.394 |
0.045 |
0.060 |
3 |
βlc |
.6 |
0.586 |
0.246 |
0.333 |
4 |
βlc |
.3 |
0.397 |
0.105 |
0.102 |
using the proposed method are closer to the true curves than the ones obtained using
the Coxph method. These results are confirmed by Figure 3.1 and Figure 3.2.
Test 2
In Test 2, a Uniform (0,5) is used for Si and τ = 0.8. Percentages of T, C and S
are 34.5%, 51.5%, and 14%, respectively, different from Test 1.
From Tables 3.3 and 3.4, it can be seen that the proposed method outperforms the
traditional Coxph method. Similarly as in Test 1, this result is confirmed by Figure
3.3 and Figure 3.4.
In summary, when the dependent censoring exists with certain magnitude, the
proposed method provides more accurate estimates of parameters and of survival
curves than the traditional Coxph method. In fact, Coxph model assumes indepen-
dent censoring.