A multiple additive regression tree analysis of three exposure measures during Hurricane Katrina |
| |
| Authors: | Andrew Curtis Bin Li Brian D Marx Jacqueline W Mills John Pine |
| |
| Institution: | 1. Visiting Associate Professor in the Department of Geography, University of Southern California, Los Angeles, United States;2. Assistant Professor in the Department of Experimental Statistics, Louisiana State University, Baton Rouge, United States;3. Professor in the Department of Experimental Statistics, Louisiana State University, Baton Rouge, United States;4. Assistant Professor in the Department of Geography, California State University, Long Beach, United States;5. Director of the Research Institute for Environment, Energy and Economics, Appalachian State University, Boone, United States. |
| |
| Abstract: | This paper analyses structural and personal exposure to Hurricane Katrina. Structural exposure is measured by flood height and building damage; personal exposure is measured by the locations of 911 calls made during the response. Using these variables, this paper characterises the geography of exposure and also demonstrates the utility of a robust analytical approach in understanding health‐related challenges to disadvantaged populations during recovery. Analysis is conducted using a contemporary statistical approach, a multiple additive regression tree (MART), which displays considerable improvement over traditional regression analysis. By using MART, the percentage of improvement in R‐squares over standard multiple linear regression ranges from about 62 to more than 100 per cent. The most revealing finding is the modelled verification that African Americans experienced disproportionate exposure in both structural and personal contexts. Given the impact of exposure to health outcomes, this finding has implications for understanding the long‐term health challenges facing this population. |
| |
| Keywords: | disaster health exposure Hurricane Katrina multiple additive regression tree recovery |
|
|
