|
Abstract:
|
In this paper, we aim to simultaneously compare multiple AUCs based on the generalized variable method (GVM) and the optimal linear combination (OLC) of multiple biomarkers. In many applications, biomarkers are frequently used to detect some symptoms or irregularities. In real diagnostic researches, sometimes it is necessary to perform multiple diagnostic tests on each individual, but how to combine the available information to increase the diagnostic accuracy is an important issue. Traditionally, researchers are used to consider one group or compare two groups at a time since the multiple group problems are highly complicated. However, in the real situation, it is likely to existing more than two groups to compare. Therefore, a comprehensive analysis to simultaneously compare the overall diagnostic accuracy of several markers deserves further exploration. In this paper, we will provide the GVM method to compare the multiple AUCs based on the combination of multiple biomarkers. Our model will assume to follow the multivariate normal distribution. The merits of the proposed method will be illustrated with numerical examples to demonstrate its advantage and application.
|
