Two Procedures for Robust Monitoring of Probability Distributions of Economic Data Streams induced by Depth Functions

摘要

Data streams (streaming data) consist of transiently observed, evolving intime, multidimensional data sequences that challenge our computational and/orinferential capabilities. In this paper we propose user friendly approaches forrobust monitoring of selected properties of unconditional and conditionaldistribution of the stream basing on depth functions. Our proposals are robustto a small fraction of outliers and/or inliers but sensitive to a regime changeof the stream at the same time. Their implementations are available in our freeR package DepthProc.