Bayesian model selection validates a biokinetic model for zirconium processing in humans

BMC Systems Biology

Table 1 Bayes factors

Inv.	$B_{H, I}^{A}$	$B_{H, I}^{p}$	$B_{H, I}^{u}$
1	7.17·10¹	7.12·10¹	1.05
2	1.15·10²	2.93·10²	3.94·10³
3	5.95·10⁴	5.23·10⁴	1.34
4	1.07·10³	2.64·10³	3.47·10¹
5	2.19·10²	4.73·10²	1.34·10²
6	4.64·10³	3.93·10³	2.38·10³
7	2.18·10²	2.30·10²	1.34·10³
8	3.75·10¹	1.28·10²	0.22
9	4.62·10²	2.32·10²	0.18
10	8.62·10²	1.16·10²	0.20
11	1.17·10⁵	1.81·10¹	2.94·10³
12	1.78·10²	5.48	1.14·10¹
13	7.19·10²	1.41·10¹	4.41
14	3.58·10¹	7.43	9.77
15	6.29·10³	2.17·10¹	1.60·10²
16	6.22·10²	1.34·10¹	1.20·10⁴
ALL	1.20·10¹¹	3.43·10⁴	4.73·10⁷

Bayes factors for the HMGU versus the ICRP model ( $B_{H, I}^{A}$ ) for investigation 1,…,16 and the complete data model (ALL) as well as the according Bayes factors for the plasma ( $B_{H, I}^{p}$ ) and urine ( $B_{H, I}^{u}$ ) data. The HMGU model is favored substantially, when $B_{H, I}^{\cdot} > 3$ and decisively, when $B_{H, I}^{\cdot} > 100$ . Also, $1 / B_{H, I}^{\cdot} = B_{I, H}^{\cdot}$ .

ISSN: 1752-0509