Visual Analytics of Spatial Scan Statistic Results

Kulldorff’s scan statistic[1] is a spatial scan statistics method for detecting and evaluating statistically-significant, spatial clusters (e.g. disease, crime, etc). The method and its software implementation – SaTScan – is used widely in an increasing number of applications including epidemiology and other research fields. Here, we abbreviate the method as SaTScan method. Many researchers have effectively applied SaTScan to small or medium size sets of geographically-referenced data (e.g. point data for cases in a city, counties within one or a few states). However, the method is sensitive to user-controlled parameter choices. Our research to address this problem prompted a broader question on the consistency of SaTScan results. This research employs visual analytics methods to (1) find and illustrate some limitations of SaTScan method, (2) facilitate tuning of SaTScan parameters to meet the needs of different categories of users, (3) enhancing the effectiveness of the method, particularly for relatively large datasets. The proposed methods are implemented in a software system called the Visual Inquiry Toolkits (VIT). We demonstrate our research by analyzing cervical cancer mortality data aggregated by county in the U.S. from 2000 to 2004.