CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Exploratory functional data analysis from depth-based neighborhoods Authors:  Antonio Elias - Universidad Carlos III de Madrid (Spain)
Raul Jimenez - Universidad Carlos III de Madrid (Spain) [presenting]
Abstract: The concept of depth has played an important role solving problems of ordering, outlier detection and clustering. We present an exploratory tool that visually provides more insights about the structure of a functional data set by the study of a simple undirected graph. To do so, we use the concept of depth-neighbourhood for defining a measure of closeness. This allows us to create a network with sample functions as nodes providing a different framework for studying a functional data set through the topology of the graph. Among others features, we show that a disconnected graph reveals the existence of clusters and how the degree of a node highlights inlier functions, outliers and groups boundaries.