Correcting inaccurate background mortality in excess hazard models through breakpoints

Table 2 Percentage of times each model was retained on the basis of its AIC

Scenario	Model	%AIC	Model	%AIC	Model	%AIC
A	1	66.70	1	66.70	1	64.78
	2	08.40	2	13.46	2	08.40
	3.1	24.90	3.2	12.55	4	26.82 (90.89,09.11)^a
B	1	04.50	1	05.82	1	04.29
	2	44.94	2	67.31	2	43.62
	3.1	50.56	3.2	26.87	4	52.09 (89.58,10.42)^a
C	1	00.60	1	01.30	1	00.60
	2	25.03	2	51.65	2	24.02
	3.1	74.37	3.2	47.05	4	75.38 (87.49,12.51)^a
D	1	24.33	1	33.63	1	23.82
	2	12.31	2	23.32	2	12.21
	3.1	63.36	3.2	43.05	4	63.97 (87.69,12.31)^a
E	1	00.00	1	00.00	1	00.00
	2	54.96	2	77.58	2	52.81
	3.1	45.04	3.2	22.42	4	47.19 (89.97,10.03)^a
F	1	13.94	1	17.85	1	13.84
	2	34.50	2	55.47	2	33.70
	3.1	51.56	3.2	26.68	4	52.46 (91.27,08.73)^a

%AIC: Percentage of times each model (between compared models) was retained according AIC
^aPercentage of times Models 3.1 and 3.2 were retained
Scenarios A to E: Proportion of subjects with x = 1 is 90%; Scenario F: Proportions of subjects with x = 1 and x = 2 are respectively 10 and 80%

ISSN: 1471-2288