Nonlinear Survival Regression Using Artificial Neural Network

Table 3

Results of concordance indexes of simulation study in testing subset (300 cases with 100 replications).

Model*	ANN	Cox Reg.
I:
20%	0.821±0.028	0.820±0.031
30%	0.819±0.027	0.820±0.030
40%	0.824±0.028	0.823±0.030
50%	0.823±0.027	0.823±0.029
60%	0.823±0.027	0.822±0.030
70%	0.825±0.027	0.822±0.032
80%	0.826±0.030	0.823±0.031
II:
20%	0.818±0.029	0.817±0.031
30%	0.817±0.034	0.815±0.046
40%	0.818±0.032	0.818±0.031
50%	0.816±0.035	0.814±0.032
60%	0.816±0.031	0.813±0.032
70%	0.814±0.033	0.811±0.031
80%	0.812±0.034	0.808±0.036
III:
20%	0.808±0.030	0.798±0.035
30%	0.799±0.033	0.790±0.033
40%	0.812±0.043	0.795±0.038
50%	0.805±0.028	0.795±0.030
60%	0.804±0.028	0.790±0.033
70%	0.805±0.028	0.791±0.034
80%	0.803±0.028	0.790±0.032
IV:
20%	0.764±0.033	0.759±0.036
30%	0.773±0.029	0.762±0.033
40%	0.777±0.030	0.759±0.035
50%	0.779±0.035	0.764±0.034
60%	0.812±0.039	0.790±0.036
70%	0.764±0.036	0.744±0.035
80%	0.766±0.035	0.741±0.037

*Four models with different rates of censoring.