Assessment of performance of survival prediction models for cancer prognosis

Table 5 Effect of training and test set sizes on the power for Model A

Training: Test	Single Group Analysis	High- versus Low-Risk Group Analysis
Training: Test	Cox Model I	Cox Model II	Log-rank test
78:19	0.3945	0.2623	0.2819
65:32	0.6000	0.4164	0.4312
49:48*	0.7630	0.5032	0.5164
32:65	0.7746	0.5226	0.5294
25:72	0.7042	0.5166	0.5232
19:78	0.563	0.4058	0.412

* From Table 4.
The 97 total patients were randomly split into a training set and a test set. The numbers of patients for the training set investigated were 78, 65, 32, 25, and 19. The values are the proportion that the estimated p-values were less than or equal to 0.05 from a total of 10,000 computations, based on 10,000 randomly splits.

ISSN: 1471-2288