general methodology

To carry out the study, our main priority is to ensure all of our data are properly formatted for our analysis. Because our data come from three different sources, a major part of this study is to correct the data prior to performing the analysis using ArcMap 9.3. In general, our tasks include creating tables from raw crime data, separating polygon using intersect features in ArcMap, combining and recalculating census statistical values in excel worksheet, performing analysis in ArcMap, as well as finalizing and retouching of maps using image editing software known as Adobe Illustrator and Corel Draw.

Preparing the data

In preparation for the data in this study, we follow a series of sequential stages including planning, followed by preparation, digitizing/transfer, editing/improvement and evaluation.

Planning is done exclusively when we come up with a specific topic for our research project. During the planning stage, we ensure sufficient and reliable data can be obtained within a reasonable amount of time to carry out the study.

Preparation is especially important in our study and is depended on the amount of raw data that are required for this study. It involves tasks such as obtaining data (transferring crime data from published report to table for importation to ArcMap), and ensuring the base maps (Local Area Boundary map, and Census Base maps) are error free for our study purpose. Also, a close overview of the available census statistical variable is carried out to ensure that we select the most meaningful numbers from census to perform the research.

Digitizing and transfer is one of most time consuming steps in our study. This includes transferring crime data and census data into ArcMap using "Join" with the relative shapefile so that the data can be displayed with its spatial reference.

Editing and improvement is carried out throughout the study when we encounter any problems during the analyzing process. For instance, when performing the "Intersect" task, some polygons are counted more than once due to the misalignment to other shapefile boundaries. Accordingly, changes in the variance setting in “Intersect” allow us to narrow the criteria when joining two different shapefiles.

Evaluation is performed continuously as decisions on the uses of different census variables are reviewed to ensure that the most representative data are selected to display our analysis.

step by step

Creating Crime Map

We are able to obtain crime statistics based on neighbourhoods in Vancouver from the Vancouver Police Department website. The data are organized in a tabular format in PDF file from the Police Department for each year. We acquire the years 2002 and 2006 data for our analysis. These data are then manually transferred to an excel spreadsheet. Differences in crime rates between the two years are also being computed prior to importing to ArcMap.

click to enlarge

This excel spreadsheet is prepared for importing into ArcMap. In addition, "Join" is performed with one of our base map "Local Area Boundaries" based on the "Area Code" of each neighbourhood to display the statistical results in a spatial manner. Two different maps are created "Distribution of Automobile Theft Cases between 2002 and 2006" and "Distribution of Break & Enter Cases between 2002 and 2006" to show the changes in Break and Enter and Motor Vehicle Theft from 2002 and 2006 in Vancouver. These maps are later imported into an image editing software for finalization.

Creating Census Map

When dealing with the Census data in this study, we take a significant amount of time to ensure all the data from census are compatible with our neighbourhood boundaries. Because we decide to use Dissemination Areas (DAs) from the census data, many efforts are spent on ensuring all the DAs are counted correctly according to where they are located within each neighbourhood. In general, we take a 6 steps approach in preparing the census data from both 2001 and 2006.

Note the problem with the DA's not lining up with the neighbourhood boundaries. Majority of the time was spend to ensure the DA's were property "formatted" to fit within each neighbourhood. click to enlarge

Step 1

First step is to select all the potential variables that we would use for this analysis from the Census 2001 and 2006 datasets. During our first stage, we select more than 10 variables for each census year (these includes average rent, average housing price). These variables are exported from the datasets and imported into a separate spreadsheet in excel. This is done to eliminate other unnecessary variables and to reduce confusion in the later stages. It should be noted that some of the variables are omitted at later stage after evaluation and review of our results due to their repetitiveness to other variables.

Table showing Census data of 2001, click to enlarge

Step 2

The excel file for Census 2001 and 2006 that we have created are later imported into ArcMap. We then join both excel files to the Census DA base map based on the "DAUID" assigned to each DA. This step ensures that the proper data can be imported and placed into the corresponding regions in the base map. As a result, we have a base map from DA that contains the selected attributes for both 2001 and 2006 census data.

Step 3

Our first goal is to ensure that each DA can only fall within one neighbourhood. Because DAs do not necessarily fall within the neighbourhood boundaries, our next step is to use the “Intersect” method to split any DAs that would intersect with the neighbourhood boundaries. This step is performed to classify each DA into their corresponding neighbourhood.  It is necessary to combine the DAs because our crime data are based on neighbourhood. Therefore, by combining the DAs into neighbourhood boundaries, it can support our comparison to the crime rates in Vancouver. In addition, because some DAs do not fall exactly within the neighbourhood boundaries, we apply the XY Tolerance (15m) in the parameters menu within the intersect method. 15m is applied because closer inspections on the datasets can reveal the majority of the DAs boundaries that are shifted approximately 15m away from their corresponding neighbourhood boundaries. The map created from the intersect method is helpful as it successfully splits some of the polygons that are located at the boundaries of two neighbourhoods. However, the intersect method only splits the polygon features physically, but the attribute values remain the same for all the new polygons. This has posed problems for our analysis because the attribute value of a polygon has been ignored when the polygon splits apart. As a result, the attribute values for the original polygons are transferred to the “split polygons”. Thus, the results from the intersect method would be inaccurate for our analysis purposes as some census values are counted twice when the DAs fall on the boundaries across two or more neighbourhoods. To remedy this problem, we manually change the attribute value by giving each of the new polygons (after splitting) their proportions of values. This is done by exporting attribute table out of ArcMap as a DBF format and importing it into excel for calculation purposes.

Intersect Method. click to enlarge

Step 4

Before exporting the attribute tables, we need to create a new field, called "New DA", to ensure the modified attribute values can be linked in the later stage based on the "New DA". At first we expect that using the "FID" in the attribute table can link the revised values at the later stage. However, a closer inspection has revealed that it is impossible to do so as the FID values are not exported along with the other attribute values. After importing the attribute table into a new excel file, we need to check over each attribute value. Our goal is to find all the polygons that are split by the intersect method because these new polygons do not have the proper attribute values to represent their new areas. The polygons that have been split into multiple polygons are easily spotted with their identical DA numbers. We then use the "Search by Attribute" method in ArcMap to locate the polygon by their DA number. Once located, we can make reference to Google Earth to ensure and determine the proportion for each polygon. Manual interpretation is done to examine the proportion for the new polygon based on land use and density of infrastructure. The new proportion is applied to the new polygon under a new field in excel called "Change". For the polygons that have not been split by the intersect method, the change value is remained at 100, while "split-polygons" are given values that amount to 100 with their corresponding polygons. After applying the appropriate proportion to each polygon, new fields for all variables are created in excel. These fields are filled with formula to automatically compute their new values based on the "Change" value. We divide the "Change" value by 100 and then multiply by the census variable to compute their equivalent new values.

Table showing the new revised census data. click to enlarge

This table was later imported back in to ArcMap and later joined with the "intersected DA Map". The join was based on the "New DA" value that we created in earlier stage. To reduce confusion in later stage, we manually deleted some of the unused field in the attirbute table.

Step 5

Because intersect is performed earlier, the spatial join from the DAs to the neighbourhood boundaries is less problematic. We first perform the "Spatial Join" task with the neighbourhood boundaries and the new DA map. "Sum" is set in the spatial join menu to ensure the values from the DAs are added to form one single value for each neighbourhood. In order for the values to add up, we ensure that the variables that we use are absolute values. Absolute and relative are the two types of data within the Census dataset. Absolute refers to the exact count within a DA, whereas relative values are normalized by the population count within the DAs. Without splitting the polygon at an earlier stage, the polygons would be counted multiple times using spatial join features when they intersect with the neighbourhood boundaries. After spatial join is performed, we are able to select different variables in the symbology properties to display various results for comparison purposes in our study.

Spatial Join performed to link sum up attribute values from DA's to neighbourhood boundaries. click to enlarge

Step 6

Reviewing and finalizing are performed as the last stage of our study in the technical aspect. We compare various maps using different variables to ensure we are able to effectively compare the results to the crime maps that we produce earlier. In addition, all maps are later exported to Corel Draw and Adobe Illustrator for editing and final retouching to remove impurities from the final products.

 

"Some areas will have larger numbers of offenders residing within them because the type of housing in such areas is more accessible to those individuals at greater risk of offending. Examples include areas of privately rented accomodation with over-representation of young single males, and transient populations"

A. Bottoms and P. Wiles (1988)