A New Method for Imputing Censored Values in Crossover Designs with Time-to-Event Outcomes Using Median Residual Life

Table 3

Percentage of bias (%) under the null and alternative hypothesis and 95% confidence interval coverage probability under the alternative hypothesis : for estimating log for our proposed method when the covariance structure is AR (1) and the distributions are exponential (E), Weibull (W), and Gamma (G).

Mean pairwise correlation
Percentage of censoring		30%			50%			30%			50%
Dist	/seq	12	24	48	12	24	48	12	24	48	12	24	48
E	Bias	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5
	Bias	0.80	1	0.6	3	3.1	2.9	1.6	1.7	1.2	3	5	4.9
	Coverage	94.5	94.9	95	97	95	94.4	95.3	95	96.5	96	95.5	97
W	Bias	^-5	^-5	^-6	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5
	Bias	0.6	0.82	0.5	2.5	2.7	2.9	1.5	1.7	1.3	2.8	4.9	5
	Coverage	95	95.1	96	96	96.7	95	96	97	96.7	97	96	94.5
G	Bias	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5	^-5
	Bias	-6.1	-5.5	-4.9	-8.2	-6.5	-5.5	-5.4	-4.1	-3.5	-7.9	-8.5	-9
	Coverage	95.1	94	95	95.3	94.2	95.1	95.5	94.4	95.2	96	95.3	95.1