Skip to main content

Advertisement

Table 3 Model selection based on Deviance Information Criterion (DIC)

From: 40 years of progress in female cancer death risk: a Bayesian spatio-temporal mapping analysis in Switzerland

Deviance Information Criterion (DIC)
Age group <55 55-74 75+ <55 55-74 75+ <55 55-74 75+ <55 55-74 75 N
Model Breast cancer Uterine cancer Cervical cancer Ovarian cancer
1: P/1re 13,430 20,328 18,161 2,140 8,521 8,327 4,300 5,459 4,167 5,417 11,887 9,404
2: NB/1re 13,462 20,373 18,216 2,149 8,542 8,345 4,312 5,473 4,179 5,430 11,912 9,432
3: P/2re 13,457 20,371 18,212 2,142 8,539 8,345 4,309 5,469 4,180 5,431 11,920 9,424
4: NBN2re 13,494 20,428 18,275 2,149 8,562 8,371 4,325 5,488 4,196 5,449 11,951 9,457
  1. Lowest DIC values per cancer site and age group are highlighted in bold face. Models 1 and 3 are Poisson regression models (P), models 2 and 4 negative binomial (NB). Models 1 and 2 have one spatially structured random effect (re), models 3 and 4 an additional, unstructured random effect