Isolating changed panels and estimating common change point after sequential detection with FDR control

Abstract

Quick detection of common changes is critical in sequential monitoring of multi-stream data where a common change is referred as a change that only occurs in a portion of panels. After a common change is detected by using a SR-SUM procedure, we compare the CUSUM and EWMA methods in terms of FDR and FNR after a common change is detected. The BH-type procedure is used for both CUSUM processes and EWMA processes at the detection time to isolate the changed panels with the control of FDR (False Discovery Rate). The numerical results show that the FNR for the CUSUM method is smaller than that for the EWMA method. Also the CUSUM method has the advantage that the common change point is conveniently estimated after isolation. To reduce the FNR, the tightened FDR control procedure is considered by using the estimated number of changed panels. Further, the estimated number of changed panels can also be used to identify whether the alarm is true or false.