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

Table 2

Type 1 error (target is 5%) and power (%) for the multiple imputation with model averaging and analysis of covariance () and our proposed method (1000 simulations).

Mean pairwise correlation
Censoring percentage			30%			50%			30%			50%
Distribution		/seq	12	24	48	12	24	48	12	24	48	12	24	48
		Method
Exp	Error		4.8	4	4.9	5	4.5	5	4.5	4.6	4.9	4.8	5	5
		MERL	4.6	4	4.8	4.9	4.1	5	3.9	4	4.4	3.8	4.1	4.5
	Power		78	80	81	69	79	80	81	79	84	78	80	81
		MERL	81	87	91	78	80	83	82	94	85	84	86	87
Weibull	Error		3.8	4	4.3	4	5	4	4.1	3	4.9	4.2	3.4	4.8
		MERL	3.5	4.1	4	4.2	4	4.3	3.8	3.1	4	4	3.5	4.2
	Power		79	80	81	72	80	81	79	80	78	81	79	80
		MERL	85	90	91	75	81	90	89	95	96	89	90	91
Gamma	Error		1.3	1.5	2.7	4.5	3.1	4.6	2.2	2.9	3.4	1.8	2.4	4.1
		MERL	1	1.2	2.1	4.1	4.1	5	1.9	2	3.4	2	2.1	3.9
	Power		77	80	79	75	78	80	81	81.9	79	81	83	82
		MERL	88	95	93	80	90	95	90	95	96	88	92	94