Table 2 Parameter estimates, model goodness of fit, model complexity, and prediction accuracy (case study II)
Whole data analysis200 CVs
ModelCovariatesaOncotype DXWhole-genome gene expression (WGGE)Estimated variance (90% posterior confidence region)Log likelihoodbEffective number of parameters (pD)Deviance information criteria (DIC)Average CV-AUC,c all luminalsAUC model in column > AUC model in row,d all luminalsAverage CV-AUC,c lymph node negativeAUC model in column > AUC model in row,d lymph node negative
Oncotype DXWhole-genome gene expressionM10M11M10M11
M9 (COV)X−59.74.4123.70.703e (0.026)0.96>0.990.689e (0.052)0.430.99
M10 (COV + ONCO)XX0.027 (0.003; 0.056)−45.34.294.90.725f (0.033)0.990.685e (0.055)>0.99
M11 (COV + WGGE)XX0.439 (0.083; 0.931)−37.79.284.60.774g (0.031)0.755f (0.039)
  • a Age and race (African American, Y/N).

  • b Estimated posterior mean of the log likelihood.

  • c Average over 200 tenfold CVs.

  • d Proportion of times that the model in column had AUC > the model in row (in 200 tenfold CVs).

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