Confounding and missing data in cost-effectiveness analysis: comparing different methods.

The AUC differences for effectiveness were highly significant for the unadjusted bootstrapped difference (-0.38), but the adjusted mean -0.04 based on predictive margins was not significant and close to zero (P-values 0.000 and 0.566, respectively), whereas the predictive mean difference based on the Bayesian inference produced a much smaller difference of -0.03 with credible interval (CrI) (-0.27, 0.20) (Table 3 ).

also in 132,142 other papers