. 2019 Feb 4;9:1218. doi: 10.1038/s41598-018-38046-9

Table 1.

Comparison of model performances on different data sets.

Data sets: organisms, radionuclides		Best-supported model (the one with lowest AICc)	Sum of Akaike weights for GE and PD models	Minimum RMSE loss for
Data sets: organisms, radionuclides		Best-supported model (the one with lowest AICc)	Sum of Akaike weights for GE and PD models	GE and PD models	ME and BE models
I. Humans, plutonium	Subject 1	PD	1.000	110.3	7706.3
	Subject 2	PD	1.000	19.3	7367.0
	Subject 3	PD	1.000	−61.6	2246.0
	Subject 4	BE	0.379	1037.2	144556.7
	Subject 5	PD	1.000	166.7	28807.9
II. Humans, strontium	Subject 1	ME	0.272	406.7	1002.1
	Subject 2	GE	0.667	248.8	4543.5
	Subject 3	ME	0.002	838.5	1132.7
III. Laboratory animals	Mouse	ME	0.493	42.4	67.6
III. Laboratory animals	Sea urchin	BE	0.549	273.5	729.4
IV. Wild animals in Fukushima nuclear accident zone, caesium	Wild boar	GE	0.943	26.6	50.3
	Black bear	ME	0.245	107.8	106.7
	Sika deer	BE	0.277	48.9	−23.9
	Ocellate spot skate	BE	0.533	−121.5	−132.1
V. Wild animals in Chernobyl nuclear accident zone, caesium	Wild boar	NA	NA	NA	NA

Instances when CPD models were favoured over discrete rate models are shown in bold font. The NA label for data set V indicates that these data were explained exclusively by physical decay of radionuclides, as described in the main text, and all tested models therefore produced equivalent fits.