IS 6833 ANALYTICS ASSIGNMENT Ying Chen, Sri Murali, John Powell, Scott Weber
Homicide - National Trend Definition: Willful (non-negligent) killing of one human being by another 1
Our Approach Study data by neighborhoods Analyze historical homicide data and develop regression model Evaluate independent variables correlated to homicide rates. Select variables to study Select model for prediction Draw conclusions about where murder will occur. Predict homicide rate for location
Regression Model Based on Historical Data Weighted Moving Average.1 – 3 previous years.2 – 2 previous years.3 – previous year.4 – current year This statistical analysis was used for the total number and an individual neighborhood basis
Regression Model - Prediction
Regression Model Conclusion We predict that 43/91 neighborhoods (47.25%) will be murder free. Prediction 1. Baden (11) 2. Jeff Vanderlou (9) 3. Midtown/Hamilton Heights/The Great Ville/ Wells- Goodfellow (5)
Demographic Based Model Study Independent Variables Correlated to Homicides Sources: Report published Bureau of Justice Statistics in report Homicide Trends in the United States, , released in November Research Paper Structural Determinants of Homicide: The Big Three, published in Journal of Quantitative Criminology in March Article National Case-Control Study of Homicide Offending and Methamphetamine Use, published Journal of Interpersonal Violence, published in June Research paper Crime is the Problem: Homicide, Acquisitive Crime, and Economic Conditions, published in Journal of Quantitative Criminology in September
Data Analysis Review Almost 90% of the offenders are males 65% of the offenders are in the age group Larger cities experienced higher number of homicides More than 2/3 rd s of homicides were by guns Homicide rates correlated to other factors Economic Conditions Educational Level Divorce Rate Drug Use
Educational Level
Poverty Level
St. Louis City Demographics Analyzed data from 2010 census for census tracts in St. Louis city Studied variables Number of males in age group Educational level Marital Status Poverty level Median Home Price
St. Louis City Homicide Data
St. Louis City - Educational Level Expressed as a percentage of male population
Divorce Rate Number of men divorced compared to number of men married
Economic Conditions Percentage of Population in Poverty
Census Tracts with High Likelihood Males aged as a percentage of population Number of men without a high school diploma Divorce Rate Among Men Percentage of Population in poverty level
Prediction Based on Historical Data Based on historical demographic information, homicide is more likely to occur in the neighborhood of Wells-Goodfellow, which is comprised of census tracts 1062 and 1063 Map of neighborhood 9 :
Conclusion Given the research on the variables we have chosen to study are not up to date, we conclude that the best predictor of murder rates in neighborhoods in St. Louis city is the regression model. We believe the variables are valid indicators of murder rates, but accurate conclusions cannot be drawn because the data is not current. Baden, with 11 predicted murders in 2012, is the most dangerous neighborhood in the city.
Assumptions and Caveats Demographic data obtained from 2010 census Census data doesn’t accurately portray current demographics Data from St. Louis Metropolitan Police Department assumed to include only homicides More recent numbers were assigned higher weights Demographic data for correlation based on single census tract
References 1: FBI definition: 2: Homicide Rates: /tables/10tbl01.xlshttp:// 2010/tables/10tbl01.xls 3: BJS report and data: 4: Structural Determinants of Homicide: The Big Three: 5: National Case-Control Study of Homicide Offending and Methamphetamine use: 6: Crime is the Problem: Homicide, Acquisitive Crime, and Economic Conditions: -live -live 7: Census data: 8: St. Louis Metropolitan Police Department Crime Statistics: 9: Map showing St. Louis City Neighborhoods and Census Tracts: