Bayesian network-based missing mechanism identification (BN-MMI) method in medical research

Table 2 The validity, simulation consistency and robustness of the BN-MMI method

Missing mechanism	Missing rate	Validity	Robustness	simulation consistency*
Missing mechanism	Missing rate	Validity	Robustness	T → M	M → T	T → O	O → W	W → O	O → M	M → O	W → M	W → T	T → W
MCAR	0.05	0.9698	0.0018**	0.0069	0.0000	0.9931	0.9838	0.0080	0.0058	0.0003	0.0065	0.0006	0.0000
	0.10	0.9727		0.0057	0.0000	0.9943	0.9847	0.0074	0.0052	0.0005	0.0053	0.0003	0.0000
	0.15	0.9686		0.0062	0.0000	0.9938	0.9829	0.0082	0.0059	0.0003	0.0075	0.0000	0.0000
	0.20	0.9697		0.0060	0.0000	0.9940	0.9831	0.0079	0.0074	0.0003	0.0067	0.0003	0.0000
MAR	0.05	0.9334	0.0235	0.0030	0.0000	0.9970	0.9889	0.0082	0.9451	0.0077	0.0011	0.0000	0.0000
	0.10	0.9794		0.0003	0.0000	0.9997	0.9889	0.0071	0.9911	0.0086	0.0004	0.0002	0.0000
	0.15	0.9790		0.0004	0.0000	0.9996	0.9874	0.0084	0.9913	0.0087	0.0001	0.0002	0.0000
	0.20	0.9825		0.0004	0.0000	0.9996	0.9882	0.0092	0.9952	0.0048	0.0004	0.0004	0.0000

^*For the sake of convince, let T stand for tumor type, M for missing fee, O for operation, and W for twoweeks
^**For the performance measures, the numbers in italic font suggested the lower they were, the better the performance was; on the contrary, the numbers in underlined bold font suggested that the higher they were, the better the performance was. Italics and underlined bold here were marked by the author, which only for ease of reading and have no special meaning

ISSN: 1472-6947