. Author manuscript; available in PMC: 2015 Sep 3.

Published in final edited form as: Psychometrika. 2014 Sep 17;80(3):811–833. doi: 10.1007/s11336-014-9413-1

Table 1.

Comparison of Cox regression, survival tree, bagging, and random survival forests in analyzing the recidivism data.

	Cox regression		Survival tree (conditional inference)		Baggin OOB Brier’s score	Random survival forests Variable importance
	Parameter estimate (SE)	Hazard ratio (SE)	Split	Groups	Baggin OOB Brier’s score	Random survival forests Variable importance
PERSONAL	.5691 (.2052)	1.7659(3642)	No	(1) AGE≤31.5;	.2142	.0222
PROPERTY	.9358 (.3509)	2.5482 (.8941)	Yes	(2) AGE>31 and PROPERTY =1;	.2142 .2178	.0222 .0117
	.9358 (.3509)	2.5482 (.8941)	Yes	(3) AGE>31 and PROPERTY = 0;	.2178	.0117
AGE	−.0667 (.0168)	.9355 (.0157)	Yes	(3) AGE>31 and PROPERTY = 0;	.1991	.0341

OOB Brier’s scores shown in the table are prediction errors of the bagging procedure without the covariate. The OOB Brier’s score with all covariates is .2123.