Table 2. Statistical properties of different correction techniques for a selected number of values for the probability of each treatment being effective.
PPV | NPV | FWER | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
K | P | Uncorrected | FWER control | FDR control (Corr) | Uncorrected | FWER control | FDR control (Corr) | Uncorrected | FWER control | FDR control (Corr) |
3 | .05 | 0.687 | 0.834 | 0.833 | 0.995 | 0.991 | 0.990 | 0.060 | 0.025 | 0.025 |
3 | .20 | 0.912 | 0.956 | 0.949 | 0.976 | 0.961 | 0.962 | 0.052 | 0.024 | 0.031 |
3 | .50 | 0.977 | 0.986 | 0.981 | 0.918 | 0.886 | 0.897 | 0.034 | 0.022 | 0.029 |
3 | .80 | 0.994 | 0.995 | 0.994 | 0.747 | 0.705 | 0.728 | 0.015 | 0.013 | 0.015 |
3 | .95 | 0.999 | 0.999 | 0.999 | 0.392 | 0.361 | 0.380 | 0.004 | 0.004 | 0.004 |
5 | .05 | 0.709 | 0.884 | 0.883 | 0.995 | 0.988 | 0.988 | 0.089 | 0.026 | 0.027 |
5 | .20 | 0.921 | 0.973 | 0.961 | 0.976 | 0.951 | 0.955 | 0.076 | 0.024 | 0.038 |
5 | .50 | 0.977 | 0.990 | 0.982 | 0.922 | 0.864 | 0.896 | 0.052 | 0.023 | 0.044 |
5 | .80 | 0.994 | 0.996 | 0.994 | 0.768 | 0.695 | 0.753 | 0.024 | 0.017 | 0.024 |
5 | .95 | 0.999 | 0.999 | 0.999 | 0.423 | 0.371 | 0.416 | 0.006 | 0.005 | 0.006 |
10 | .05 | 0.747 | 0.935 | 0.932 | 0.995 | 0.985 | 0.984 | 0.140 | 0.025 | 0.027 |
10 | .20 | 0.927 | 0.984 | 0.967 | 0.977 | 0.937 | 0.947 | 0.126 | 0.026 | 0.055 |
10 | .50 | 0.978 | 0.994 | 0.981 | 0.922 | 0.823 | 0.890 | 0.089 | 0.025 | 0.076 |
10 | .80 | 0.994 | 0.997 | 0.994 | 0.788 | 0.659 | 0.776 | 0.043 | 0.022 | 0.043 |
10 | .95 | 0.999 | 0.999 | 0.999 | 0.474 | 0.378 | 0.470 | 0.012 | 0.010 | 0.012 |
Abbreviations: FDR, false-discovery rate; FWER, family-wise error rate; K, number of experimental arms; NPV, negative predictive value; P, probability of each experimental arm being effective; PPV, positive predictive value.