TABLE 3.
Estimated Coverage of the 95% CIs Based on Binormal and Empirical AUC, and the Relative Bias of the Estimated Variance of the Binormal AUC for the 10,000 Bigamma Datasets Generated for Each Scenario
| Sample Size | Number of Categories | True Bigamma AUC | κ = 0.1 (Highly Non-binormal)
|
κ = 0.33
|
κ = 3 (Binormal-like)
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Relative Bias of the Estimated Variance | Estimated Coverage of a Binormal 95% CI | # | Estimated Coverage of a Non-parametric 95% CI for Continuous Data | Relative Bias of the Estimated Variance | Estimated Coverage of a Binormal 95% CI | # | Estimated Coverage of a Non-parametric 95% CI for Continuous Data | Relative Bias of the Estimated Variance | Estimated Coverage of a Binormal 95% CI | # | Estimated Coverage of a Non-parametric 95% CI for Continuous Data | |||
| 100:100 | 10 | 0.6 | 0.00 | 94.7% | 10,000 | 94.9% | −0.09 | 94.1% | 10,000 | 94.9% | −0.14 | 94.2% | 10,000 | 94.8% |
| 0.7 | 0.08 | 95.5% | 10,000 | 94.9% | 0.05 | 96.1% | 10,000 | 94.8% | −0.02 | 95.6% | 10,000 | 94.3% | ||
| 0.8 | 0.10 | 96.0% | 10,000 | 94.2% | 0.05 | 96.3% | 10,000 | 94.6% | −0.04 | 95.6% | 9996 | 94.6% | ||
| 0.9 | 0.00 | 96.2% | 9998 | 93.8% | 0.01 | 96.2% | 10,000 | 93.4% | 0.49 | 95.6% | 9981 | 93.6% | ||
| 50:50 | 10 | 0.6 | 0.03 | 95.1% | 10,000 | 94.8% | −0.06 | 94.6% | 10,000 | 94.4% | −0.09 | 94.8% | 9999 | 94.9% |
| 0.7 | 0.07 | 95.3% | 10,000 | 94.6% | 0.01 | 95.1% | 10,000 | 94.4% | −0.02 | 95.0% | 10,000 | 94.6% | ||
| 0.8 | 0.08 | 94.9% | 10,000 | 93.7% | 0.03 | 95.2% | 10,000 | 93.8% | 0.00 | 94.9% | 9998 | 93.4% | ||
| 0.9 | 0.07 | 94.8% | 9869 | 91.8% | 0.05 | 95.4% | 9948 | 91.7% | 0.09 | 94.8% | 9992 | 92.2% | ||
| 100:100 | 5 | 0.6 | 0.02 | 97.1% | 10,000 | 94.9% | 0.00 | 96.9% | 10,000 | 94.9% | 0.00 | 96.8% | 10,000 | 94.8% |
| 0.7 | 0.03 | 97.1% | 10,000 | 94.9% | 0.04 | 97.5% | 10,000 | 94.8% | −0.01 | 97.0% | 10,000 | 94.3% | ||
| 0.8 | 0.02 | 97.8% | 9993 | 94.2% | 0.01 | 97.6% | 10,000 | 94.6% | 0.01 | 96.9% | 10,000 | 94.6% | ||
| 0.9 | 153.69 | 99.7% | 8994 | 93.8% | 89.56 | 100.0% | 9167 | 93.4% | 3.34 | 97.1% | 9960 | 93.6% | ||
| 50:50 | 5 | 0.6 | 0.02 | 96.9% | 10,000 | 94.8% | −0.01 | 96.5% | 10,000 | 94.4% | 0.01 | 96.9% | 10,000 | 94.9% |
| 0.7 | 0.03 | 97.1% | 10,000 | 94.6% | 0.00 | 96.7% | 10,000 | 94.4% | −0.03 | 96.3% | 10,000 | 94.6% | ||
| 0.8 | 0.13 | 96.7% | 9957 | 93.7% | 0.06 | 96.9% | 9996 | 93.8% | 0.02 | 96.4% | 10,000 | 93.4% | ||
| 0.9 | 129.61 | 98.2% | 8368 | 91.8% | 86.16 | 99.9% | 9104 | 91.7% | 18.33 | 96.5% | 9820 | 92.2% | ||
AUC, area under the ROC curve; CI, confidence interval; ROC, receiver operating characteristic.