Table 4

Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for different cut-offs for probability of non-response at 1 year based on the predicted probability of non-response for each model and the average predicted probability across the five models

ModelCut-offSensitivitySpecificityPPVNPV
0.30.860.480.620.80
LR0.50.650.620.670.67
0.70.370.910.800.60
0.30.910.360.580.80
RF0.50.740.610.650.71
0.70.420.860.750.60
0.30.930.270.560.80
SVML0.50.670.680.670.68
0.70.190.930.730.54
0.30.930.300.560.81
SVMR0.50.700.610.640.68
0.70.370.950.890.61
0.30.810.480.600.72
ANN0.50.650.680.670.67
0.70.330.840.670.56
Average0.30.930.300.560.85
prediction across0.50.700.730.710.71
the five models0.70.300.950.870.58
  • ANN, artificial neural networks; LR, logistic regression; RF, random forest.