A New Model Selection Metric for Biomarker Detection Algorithms and Tools
Table 5
Simulation result when Pc = 0.5.
Cutoff
Sen
Spe
PPV
NPV
St
N_random
N_screening
L
Cost
F
Gmeans
7
0.996
0.909
0.916
0.996
0.335
140.189
281.951
0.088
16838410
7.101
0.952
8
0.99
0.952
0.954
0.99
0.343
133.554
270.239
0.056
16057748
6.851
0.971
9
0.977
0.977
0.977
0.977
0.349
129.683
265.895
0.045
15627214
6.73
0.977
10
0.952
0.99
0.99
0.954
0.351
127.701
268.749
0.056
15457568
6.833
0.971
11
0.909
0.996
0.996
0.916
0.353
126.742
279.444
0.088
15468656
7.06
0.951
12
0.841
0.999
0.998
0.863
0.353
126.148
300.435
0.139
15619148
7.293
0.917
13
0.747
1
0.999
0.798
0.353
126.034
337.981
0.202
15983172
7.494
0.864
14
0.63
1
1
0.73
0.354
126.006
400.571
0.27
16606314
7.658
0.794
15
0.5
1
1
0.667
0.354
126.001
505.193
0.333
17651994
7.794
0.707
Note: Sen: sensitivity; Spe: specificity; St: standardized treatment effect; N_random: randomized sample size; N_screening: screening sample size; L: predictability loss of M; cost: estimated total cost in an M-based drug clinical trial.
