B0624
Title: A clustering approach for random intervals based on an overlapping measure
Authors: Ana Belen Ramos-Guajardo - University of Oviedo (Spain) [presenting]
Abstract: A new method for clustering random intervals is proposed. It is based on the degree of overlap between intervals according to the Szymkiewicz-Simpson coefficient. In this sense, two random intervals can be grouped in the same cluster whenever the overlapping measure between their expected values is assumed to be greater than or equal to a specific degree. To verify such an assumption, a two-sample overlapping bootstrap test can be carried out on each pair of random intervals, leading to a p-value matrix. Each p-value can be interpreted as a kind of similarity amongst the random intervals, so that the greater the p-value, the higher the degree of intersection of two expectations and, hence, the higher the similarity between the corresponding random intervals. Finally, a hierarchical clustering algorithm for grouping random intervals is proposed, analyzing its behaviour by means of simulation studies and by applying it to a real-life situation.