Table 5. Likelihood of causal association between low 25-OHD and colorectal cancer is compared with the reverse causal hypothesis (proposing CRC leads to lower 25-OHD), on the complete dataset and for a range of parameter settings.
setting | |||||
S1 | S2 | S3 | S4 | S5 | |
DIC causal | −10,480 | −11,074 | −4,242 | −2,203 | 3,004 |
DIC reverse | −4,161 | −358 | −3,661 | 1,115 | 5,214 |
DIC difference | 6,319 | 10,715 | 580 | 3,318 | 2,210 |
mean DIC | −7,321 | −5,716 | −3,952 | −544 | 4,109 |
Deviance information criterion (DIC) score differences between models are shown; positive values indicate that causal association is more likely. Mean DIC is calculated as the average DIC for all causal and reverse causal models considered for any given parameter setting (smaller values indicate better models). Large positive DIC differences provide overwhelming evidence for a direct causal relation between low 25-OHD and colorectal cancer. Details on DIC components are in Table S2. Digits after decimal point have been omitted from the table.
Settings: S1: precx = 1000, precxt = 1000, precy = 0.1; S2: precx = 100, precxt = 100, precy = 100; S3: precx = 1000, precxt = 1000, precy = 10; S4: precx = 100,precxt = 100, precy = 200; S5: precx = 20, precxt = 20, precy = 200.