Comparison Online Monitoring Method of Correlated High-Dimensional Data Streams

In this article we consider four charts (max and sum of cumulative sum, higher criticism statistic and goodness-of-fit test statistic) for monitoring related high dimensional data streams to find the alarming time as quickly as possible after the mean shift and find the ideal chart in different condition. We use the robust of different one-side statistics, , , to find the best chart in sparse case and dense case. From the results analysis, we can gain that the goodness-of-fit test has the best efficient from balancing the power and robust in theory.