Table 4.

Results of simulations designed to emulate hazard ratio estimation for energy in relation to breast cancer in the Women's Health Initiative, using 500 simulated data sets. The mean bias, empirical standard deviation (SD), root mean squared error (RSME), type I error (α) for β = 0, and estimated 95% coverage probability (CP) for β = 1.25 are given.

β = 0	Bias	SD	RMSE	α	β = 1.25	Bias	SD	RMSE	CP
True	0.004	0.296	0.296	0.046	True	0.006	0.302	0.302	95.4
Naive	−0.008	0.105	0.105	0.056	Naive	−1.181	0.106	1.189	0.0
RC	−0.005	0.709	0.709	0.050	RC	−0.006	0.691	0.691	96.2
RRC	−0.005	0.709	0.709	0.050	RRC	−0.006	0.690	0.691	96.2
CS	−0.052	1.507	1.507	0.052	CS	−0.318	1.510	1.543	93.1
NP	0.014	1.528	1.528	0.042	NP	−0.217	1.471	1.487	95.0

True: Cox regression with true X; Naive: Cox regression with unadjusted Q; RC: ordinary regression calibration; RRC: risk set regression calibration; CS: conditional score; NP: nonparametric corrected score.