Table 3.
Stomach | Liver | Lung | |
gen | M/F | M/F | M/F |
φmb | 11.89 | 36.90 | 13.60 |
ca Dφ > 0 | 1311 | 231 | 506 |
βc | 0.46 | 1.43 | 0.44 |
σ | 119.96 | 290.09 | 37.92 |
τ | 82.67 | 76.82 | 4.47 |
LRT2p-transd | 15.68 | 16.76 | 0.34 |
LRT2p-lin | 19.64 | 27.64 | 14.73 |
LRT2p-con | 21.19 | 34.87 | 16.04 |
ERR0.025,φ | 0.39 (0.08, 0.86) | 1.10 (0.26, 2.37) | 0.86 (0.14, 3.34) |
ERR1,φ | 0.46 (0.11, 0.94) | 1.43 (0.48, 2.95) | 0.88 (0.12, 3.36) |
Dmaxe | 0.01 | 0.01 | 0.23 |
ERRDmax,φ | 0.54 (0.17, 1.04) | 1.41 (0.50, 2.86) | 3.23 (0.09, 9.39) |
Dev | 1661.53 | 1234.35 | 1038.66 |
Chi-sq | 3408.01 | 2447.95 | 2770.40 |
df | 2990 | 2990 | 2990 |
Rf | 262.62 | 204.11 | 86.39 |
Excess | 390.00 | 126.36 | 224.43 |
Colon | Uterus | All Solid | |
gen | M/F | F | M/F |
φmb | 20.28 | 26.91 | 43.86 |
Dφ > 0 | 184 | 127 | 365 |
βc | -0.74 | 0.47 | -0.62 |
σ | 11.32 | 198.10 | 13.10 |
τ | 2.66 | 81.12 | 2.32 |
LRT2p-transd | 1.32 | 1.83 | 1.14 |
LRT2p-lin | 3.01 | 6.31 | 12.90 |
LRT2p-con | 3.54 | 6.31 | 17.93 |
ERR0.025,φ | 0.25 (-0.44, 1.46) | 0.66 (-0.10, 2.44) | 0.27 (0.05, 0.54) |
ERR1,φ | 0.05 (-0.53, 1.15) | 0.47 (-0.32, 2.12) | 0.27 (-0.04, 0.66) |
Dmaxe | 0.32 | 0.01 | 0.35 |
ERRDmax,φ | 1.31 (-0.71, 1.51) | 0.90 (-0.17, 2.96) | 1.59 (0.03, 2.54) |
Dev | 716.45 | 632.66 | 2120.13 |
Chi-sq | 3339.24 | 1047.11 | 2759.48 |
df | 2990 | 1511 | 2990 |
Rf | -14.36 | 425.41 | -19.98 |
Excess | 29.34 | 45.89 | 74.59 |
a 0 – 20 mSv subcohort, Excess Relative Risk ERR = βDφ + σDφ with τ ≥ 0 and β,σ,τ constrained by 1 + ERR ≥ 0 for all data cells; for Dφ see Table 1. The transient model is nested by β = 0; the linear model is nested by σ = 0 or by τ = 0; the control model is nested by β = σ = 0.
b chosen in the range 5 – 44 years. to minimise two-phase model Deviance; where italicised φm minimises Dev subject to ERR0.025,φ ≥ 0, a weaker condition than β, σ ≥ 0.
c At φm, β, σ and τ are the fitted Maximum Likelihood Estimate parameter values.
d At φm, the likelihood ratio tests LRT2p-trans and LRT2p-lin compare the two-phase with the transient and linear models (χ2 on 1 df); and LRT2p-con compares the two-phase model with the control model (χ2 on 2 df)
e At φm, Dmax is the value of D giving a local maximum of ERRD,φ
f R = 1 + σ/β
Other definitions as in Table 2