Skip to main content
. 2020 May 2;27(5):747–756. doi: 10.1093/jamia/ocaa023

Table 3.

The experiment results for the myocardial infarction (Edin), the cancer biomarker (CA), and the length of hospitalization (THA) datasets. The evaluation metric is the averaged full area under the receiver operating characteristic curve (AUC) among N sites, for 30 trials. The Pearson Correlation Coefficient (PCC) was computed to evaluate the linear correlation between 2 methods. Finally, the alpha in the 2-sample t-test was 0.05, and the p-values larger than 0.05 (shown in bold italic) indicate no statistically significant difference between the AUC results of EXPLORER and ExplorerChain

EXPLORER
ExplorerChain
Correlation
Two-Sample t-Test
Dataset N Mean AUC Standard Deviation Mean AUC Standard Deviation PCC Delta Test Statistics P-value
Edin 2 0.965 0.013 0.965 0.013 0.999 0.000 −1.559 0.130
4 0.962 0.010 0.960 0.011 0.867 0.000 1.868 0.072
8 0.957 0.014 0.954 0.015 0.906 0.002 1.371 0.181
CA 2 0.893 0.054 0.891 0.055 0.977 0.000 1.106 0.278
4 0.862 0.075 0.853 0.078 0.932 0.000 1.694 0.101
8 0.892 0.060 0.876 0.071 0.746 0.000 1.811 0.080
THA 2 0.734 0.035 0.733 0.036 0.995 0.000 1.622 0.116
4 0.738 0.047 0.735 0.047 0.975 0.000 1.529 0.137
8 0.718 0.040 0.712 0.040 0.909 0.000 1.878 0.070

Abbreviations: AUC, area under the receiver operating characteristic curve; CA, cancer biomarker; PCC, Pearson correlation coefficient; THA, total hip arthroplasty.