—Lixin Huang, Advanced Transportation Engineering Consultants
View Presentation [PDF]
Analyzing people's behaviors in hurricane evacuation deals with peoples' responses to the threat of an oncoming hurricane. For evacuation, there are two possible responses for people who live in a threatened area. One is to leave their homes and the other is to stay. For people leaving their homes, i.e. evacuees, people's behavior may determine how far they could go. There are many factors affecting peoples' responses, such as socioeconomic and demographic data. Behavioral analysis will provide information on how these factors affect peoples' responses to a hurricane. The data used for behavioral analysis is Hurricane Katrina evacuation survey data funded by the National Science Foundation (NSF). The original survey data contains a comprehensive list of items. These items include households' experiences during Hurricane Katrina, such as the way households assess their risk level, households' decision-making processes, promoting and constraining factors while making evacuation decisions, hurricane-related knowledge, attitudes before and after Hurricane Katrina, etc. This paper will only focus on how the socioeconomic and demographic data affect people's behaviors. These data items include education, income, occupation, marital status, household size, etc. The survey was done by interviewing the randomly selected household. The survey data contains the longitude and latitude of each interviewee's home location, and the destination city and state information of each evacuee. The destination city and state information can be converted to longitude and latitude data by using geocoding method. The longitude and latitude data makes analyzing the people's behaviors in GIS possible. New spatial statistics tools, Ordinary Least Squares Regression and Geographically Weighted Regression, were introduced in ArcGIS 9.3. And Geographically Weighted Regression is one of several spatial regression techniques increasingly used in geography and other disciplines. With both tools, behavioral analysis in hurricane evacuation could be done against the socioeconomic and demographic data to find out which factors play important roles. Ordinary Least Squares Regression will be used first as the initial analysis to find out the relationship between dependent variable, e.g. the evacuation distance, and the explanatory variables (socioeconomic and demographic data). It may fail to produce the results due to the changes of relationship being analyzed across the study areas. So Geographically Weighted Regression will then be used to do the analysis to overcome the limitation of Ordinary Least Squares Regression.