Variable selection under multiple imputation using the bootstrap in a prognostic study

Table 3 The performance of methods 1 to 4 at different levels of proportions of selected indicators

	MI†				B†				MI100+BI†				MI10+BI†
		c-index		slope		c-index		slope		c-index		slope		c-index		slope
Threshold	n	AP	BC	BC	n	AP	BC	BC	n	AP	BC	BC	n	AP	BC	BC
90%	5	0.74	0.71	0.86	6	0.74	0.71	0.85	3	0.74	0.71	0.86	4	0.72	0.71	0.85
80%	8	0.76	0.71	0.85	8	0.76	0.72	0.83	5	0.74	0.71	0.84	5	0.74	0.71	0.85
70%	10	0.76	0.71	0.77	12	0.78	0.72	0.80	11	0.77	0.72	0.79	11	0.77	0.72	0.80
60%	13	0.77	0.72	0.70	18	0.79	0.71	0.75	27	0.79	0.70	0.67	26	0.79	0.70	0.64
0% (full model)	31	0.80	0.70	0.65	31	0.79	0.69	0.65	31	0.80	0.70	0.64	31	0.80	0.70	0.65

ISSN: 1471-2288