Table 4 Parameter estimates, model goodness of fit, model complexity, and prediction accuracy (case study IV)
Whole data analysis200 CVs
ModelsModels componentsEstimated variance (90% posterior confidence region)Log likelihoodeEffective number of parameters (pD)Deviance information criteria (DIC)Average CV-AUCfProportion of times model in column had AUC > model in row
COVaMETHbWGGEcMETH × WGGEdMETHWGGEMETH × WGGECOV + METH + WGGECOV + METH × WGGE
COVX−85.76.4177.90.724g (0.001)>0.99>0.99
COV + METH + WGGEXXX0.162 (0.075; 0.440)0.220 (0.090; 0.690)−73.917.6165.40.754h (0.004)0.40
COV + METH × WGGEXXXX0.101 (0.046; 0.272)0.138 (0.055; 0.474)0.101 (0.044; 0.329)−69.920.2159.90.753h (0.005)
  • a Age: African American Y/N; lobular (Y/N); and tumor subtype.

  • b Methylation.

  • c Whole-genome RNA-seq.

  • d Methylation-by-WGGE.

  • e Estimated posterior mean of the log likelihood.

  • f Average over 200 tenfold CVs.

  • g,h The same letter indicates that the models are no different (empirical P < 0.05).