Skip to main content
. 2012 Mar 22;51(2):165–178. doi: 10.1007/s00411-012-0410-4

Table 1.

For both biological endpoints, the preferable final non-nested models are shown with related final deviances (dev), difference in final deviances (Δdev) with respect to the model with the smallest deviance, number of model parameters (N par), AIC-values, difference in AIC-values (ΔAIC) with respect to the model with the smallest AIC-value, and Akaike weights

dev Δdev N par AIC ΔAIC Weight
CVD (ICD-9 430–429)
 ERR-LNT model [#1] 3569.51 3.46 22 3613.51 1.46 0.2628
 ERR-quadratic model [#2] 3570.14 4.09 22 3614.14 2.09 0.1918
 ERR-step model [#6], D th = 0.62 Gy 3566.05 0 23 3612.05 0 0.5454
 Preston’s ERR-LNT model 3599.58 33.53 30 3659.58 47.53
Cardiovascular diseases (390–429, 440–459)
 EAR-LNT modela [#1] 3693.73 0 17 3727.73 0 0.3619
 ERR-quadratic modela [#2] 3694.05 0.32 17 3728.05 0.32 0.1918
 EAR-threshold model [#5], D th = 2.0 Gy 3695.0 1.27 17 3729.0 1.27 0.1379
 EAR-step model [#6], D th = 2.19 Gy 3695.66 1.93 17 3729.66 1.93 0.3084
 Preston’s ERR-LNT model 3709.71 15.98 30 3769.71 41.98

As a comparison, the values are also shown for Preston’s ERR-LNT models. Note that for cerebrovascular disease the three preferable models are ERR models; for cardiovascular diseases, the four preferable non-nested models are EAR models. The numbers in brackets refer to the eleven dose–responses depicted in Fig. 1

aContains an age-dependent dose-effect modifier