Table 1 The out-of-sample metrics for the models (TCM, TCM-CRP and JM) and the baseline model NM with both of the prediction horizons: means of coverage probability (CP), mean absolute error (MAE) and root mean squared error (RMSE).

From: Prediction of leukocyte counts during paediatric acute lymphoblastic leukaemia maintenance therapy

 

2 weeks

4 weeks

TCM

TCM-CRP

JM

NM

TCM

TCM-CRP

JM

NM

\({\overline{{\rm{CP}}}}_{50}\)

0.457 (0.13)

0.442 (0.13)

0.421 (0.13)

0.514 (0.15)

0.443 (0.15)

0.428 (0.14)

0.402 (0.15)

0.507 (0.15)

\({\overline{{\rm{CP}}}}_{90}\)

0.795 (0.11)

0.787 (0.11)

0.761 (0.14)

0.873 (0.09)

0.767 (0.13)

0.759 (0.13)

0.722 (0.17)

0.863 (0.09)

\(\overline{{\rm{MAE}}}\)

0.860 (0.33)

0.870 (0.35)

0.964 (0.42)

0.986 (0.43)

0.896 (0.34)

0.914 (0.38)

1.016 (0.45)

1.001 (0.44)

\(\overline{{\rm{RMSE}}}\)

1.232 (0.51)

1.244 (0.53)

1.387 (0.66)

1.308 (0.57)

1.278 (0.55)

1.309 (0.64)

1.430 (0.68)

1.326 (0.59)

  1. Standard deviations are in parentheses. Similar means are obtained if the metrics are computed modelwise without considering the patients separately.