. Author manuscript; available in PMC: 2018 Jan 1.

Published in final edited form as: J R Stat Soc Ser C Appl Stat. 2016 Jun 27;66(2):313–328. doi: 10.1111/rssc.12164

Table 4.

Estimation and testing of α₁ (truth is 0.5): heavy censoring dependent on Y.

	Method	Bias	SD	SE	MSE	Type 1 error	Power
n = 100	No Censoring	0.0060	0.2709	0.2624	0.06892	0.052	0.524
	Complete Case	−0.3810	0.8527	0.8286	0.8317	0.091	0.098
	Single Imputation¹	0.1001	0.5021	0.4891
	Multiple Imputation	0.1011	0.5216	0.5314	0.2926	0.055	0.152
	Deletion Thresholding	0.0300	0.2476	0.2515	0.0641	0.059	0.129
	Complete Thresholding	−0.1049	0.2059	0.2045	0.0528	0.060	0.094
	Reverse Survival²					0.062	0.227

n = 250	No Censoring	0.0001	0.1605	0.1605	0.0258	0.047	0.872
	Complete Case	−0.3256	0.5357	0.5186	0.3750	0.110	0.157
	Single Imputation¹	0.0732	0.3896	0.2837
	Multiple Imputation	0.0999	0.3659	0.3319	0.1201	0.090	0.240
	Deletion Thresholding	0.1399	0.1777	0.1868	0.0545	0.055	0.247
	Complete Thresholding	1.3124	0.1643	0.1633	1.7491	0.054	0.151
	Reverse Survival²					0.051	0.548

: MSE, type I error and power not reported due to invalid SE for single imputation

: bias, SD and SE not reported due to non-interpretability of estimate from reverse survival regression