Master of Science (MS)
Geography and Anthropology
Because homicides are rare events, criminologists must often deal with the Small Population Problem, which creates unreliable homicide rates based on arbitrarily delineated census tracts of low population. These rates lead to violations in several assumptions required in statistical analysis. This study proposes the Regionally Constrained Agglomerative Clustering and Partitioning (REDCAP) method to mitigate the Modifiable Areal Unit Problem and solve the Small Population Problem by constructing new, larger regions with sufficient minimum populations for homicide rate calculation. This method is used for a case study of New Orleans, Louisiana, to test the relationship between concentrated disadvantage and homicide. Ordinary Least Squares and Geographically Weighted Regressions are conducted with the data both before and after the REDCAP operation. Results for the standard census tract layer show a weak and insignificant relationship between concentrated disadvantage and homicide because of extremely unreliable rate estimates. After the REDCAP operation, variables show a more normal distribution and reduced variability; moreover, regression results confirm a strong and positive relationship between concentrated disadvantage and homicide. This study shows viability for REDCAP as a regionalization method for further studies on violent crime, namely its ability to provide more stable data for improved reliability in crime rate calculations. Additionally, this study provides implications for public policy, specifically social cohesion and efficacy policies, including community-oriented policing.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Robert, Lawrence Keenan, "Constructing geographic areas for homicide research : a case study of New Orleans" (2013). LSU Master's Theses. 226.