A APPENDIX SUPPLEMENTARY TO NUMERICAL RESULTS 12

This is the appendix of the paper "Detecting Abrupt Changes in Sequential Pairwise Comparison It contains two parts: 1. Appendix A for some supplements to numerical results in Sections 4 and 5. 2. A.1 Wild binary segmentation based on likelihood It is based on the so-called CUSUM statistics. It is known that Binary Segmentation is consistent but not optimal (V enkatraman (1992)). Algorithm 3 shows the general framework of WBS algorithm. To get a rough sense of the number and locations of change points, we check the paths of the logarithm of generalized likelihood ratio statistics, which are shown in Figure 2. It should be noted that although In this subsection, we apply the potential competitor, the likelihood-based WBS method (i.e. Appendix A.1, WBS has another tuning parameter Here, we use samples with odd time indices as training data and even time indices as test data.