#metrocrime This repository contains the final paper and supporting code for our project for MATH-658 Survey Sampling at Georgetown University, given summer 2015 by Dr. Tommy Wright of the U.S. Census Bureau. We reproduce below the abstract from our paper, "Estimating Average Levels of Crime Reported within the Vicinity of a Metro Exit."

We utilize data from the Metropolitan Police Department (MPD) online database to create a finite population of property crimes occurring in the metropolitan D.C. area from January 1, 2014 through July 1, 2015. We employ clustering algorithms in the statistical package R to create 39 unique centroids defined by an individual or a cluster of Metro station exits and assign crimes to these locations. We then stratify this finite population by D.C.’s eight wards and employ Exact Optimal Allocation to construct a sample from each ward. Finally, we use statistical techniques to infer our primary parameter of interest: the average number of property crimes occurring within a 150 meter radius of a given metro exit on a weekend. We ultimately arrive at a 95% confidence interval of 0.2234 to 0.3247 property crimes per station exit per weekend.

GIS data were obtained from District of Columbia Open Data and crime data from the Metropolitan Police Department. We include our data set for reproducibility. We rely on the R packages dplyr, geosphere, ggplot2, lubridate, maptools, rgdal, rgeos, and sp, and are grateful to their creators and maintainers.