In the last decades, multiple-testing problems have received much attention. Many different methods have been proposed in order to deal with this relevant issue. Most of them are focused on controlling some weak version of the Type I error such as the False Discovery Rate. Type II error is frequently forgotten. In this work, the multitesting problem is treated from a diagnostic test approach in which the p-values play the role of the studied predictive marker. In this context, the receiver operating characteristic, ROC, curve is estimated. Several Monte Carlo simulations help for a better understanding of the problem. Finally, a real dataset studying the relationship between atosomal CpG sites and characteristic of hepatocellular carcinoma is considered.
Type of Publication:
Eduardo Gil, Eva Gil, Juan Gil, María Ángeles Gil
The Mathematics of the Uncertain. A tribute to Pedro Gil