Detecting high incidence by using variable scan radius

This research aims at detecting spatiotemporal clustering with increased mean. Scan statistics are popular methods for spatiotemporal surveillance. Several likelihood-ratio (LR) and exponentially weighted moving average (EWMA) based scan statistics have been studied for the scenarios with known or unknown size of shifted coverage. However, the existing EWMA-based methods applying fixed radii may not be efficient to detect the cluster with unknown shifted coverage. This paper proposed an EWMA-based scan statistic with variable scan radii to detect clustering instead. The proposed statistic weights the observations by distance in each circular scan window and uses the EWMA technique across the temporal axis. Comparing to the LR-based scan statistic with variable scan radii, the proposed method can be more sensitive when clusters occur at the early stage. The proposed method would have advantage in solving practical problems with unknown size of shift coverage.