 |
 |
|
|||||||||||||
|
METHODOLOGY - Park Crime My method for investigating the context of crime
for my case study parks 1. Cleaning up the crime report database The database I used includes all crime reports made in the city of Providence between December 1999 and August 2001. The database was originally received in Access format from the Providence Police Department through the Department of Administration. Crime reports are usually calls for service (incidents where the police were called) that resulted in an incident report; yet they can also be incidents initiated by police investigations, such as drug raids. Therefore, the database does not give a full account of all the crime occurrences in the city - it only shows those occurrences that were recorded as actual reports. The fields included in the database were: The database initially held approximately 150,000 records. However, many of these records were classified in crime categories that were irrelevant to the scope of my study, and other records were missing important information or contained errors that would make them inappropriate for mapping. To pare down the database, I first made a list
of all the types of crime that I considered unsuitable for a study focusing
on crime in parks. For example, I knew that categories such as 'Fraud'
or 'Embezzlement' would have no relevancy for understanding criminal trends
in a park. The next set of adjustments to the database involved
ensuring that all the street addresses were in a format that would render
them suitable for mapping in GIS. This meant that each record had to have
a specific street address. This requirement precluded the inclusion of:
I wanted to map each crime category from my pared-down
database separately. This meant that I had to make each crime category
into its own table. This simply entailed sorting the database by the crime
code, and then copying and pasting each crime category's records into
separate tables. 5/5/00 002 245 ALLENS 245 Allens Ave. had three different instances of an 'abandoned vehicle' on different dates. The issue here is that if this table was mapped in Arcview, the program would not automatically count up the three violations, and show this as the total number of abandoned vehicle violations for this address. Instead, it would recognize each record separately, and simply map them on top of one another, not allowing a concise understanding of the mapped information for the specific address. In light of this, I decided to reorganize the tables so that there would be one record for each address, which showed the total number of that type of violation at that address, and this would consequently be reflected in the maps. I used the 'subtotal' function in Excel, which inserted a subtotal each time it recognized that the address field had changed. The result was the following: 002 245 ALLENS 3 The same was applied to all the crime category tables. 2. Preparing the groundwork
for parcel-level mapping of the database
Without a master platlot-address table, I could not hope to comprehensively parcel map the crime database for the entire city. I knew that I could achieve partial success - some of the addresses for the crime records would match with those in the Arcview parcel map and would therefore be mapped - yet I had no way of knowing what the matching success rate would be. I had to find a way to at least improve my chances of a decent matching rate for the citywide mapping. After that, I knew that once I had determined specific areas around my case study parks to focus on, I could use the hard-copy platlot maps to assign platlot numbers to those records that were going unmapped. I had noticed that the addresses in the Arcview parcel map were inconsistent in their spelling and the type of identifiers they used. For example, Devereaux Street was sometimes spelled Devereux Street, and other times as Devereux St. If a record in the crime database had '11 Devereux St.', while the record for that address in the parcel map said '11 Devereaux St.', then the crime record would not be matched to the parcel to which it belonged, and therefore not be mapped. I decided that I would make a new table in Excel, taking just the platlot and address fields of the parcel-map table, reconfigure all the addresses into a common format, and then apply this same format to all the addresses in the crime database. This way, I could at least ensure that all the addresses in the parcel map would match with the corresponding addresses in the crime database. In making this 'Platlot-Address' table, I split the address field into three parts - the street number, the street name, and the street identifier - and then removed the street identifier to aid simplicity. For example, /42 John St./, became / 42 / John/ St/, and then the /St./ column was removed. To handle streets that have the same street names, i.e. Park St., Park Ave., Park Lane, etc., where removing the street identifier would create confusion as to which street was actually being referred to, I made one of the streets the 'main' street by removing the identifier, and left the other identifiers alone. Therefore Park St. became just 'Park', and the others were put into a common format with the street identifier, i.e. Park Avenue, Park Lane, etc. I also had to ensure that all the street names were spelled in the same way. Thus, all the 'Devereaux' had to be spelled the same, and names with spacing in them, such as 'De Soto', were held to a common format. Even when this was arbitrarily done, the important thing was keeping the format consistent throughout. I then went to my crime report database and applied this address format to all the records. This involved splitting the address field into its respective components - number, name, identifier, seeking out the streets with duplicate street names and putting the identifiers in with the street names, deleting all the other identifiers, and then ensuring that all the street names were correctly spelled and in exactly the same format as in the Platlot-Address table. 3. Parcel mapping the database using GIS Arcview First, the reformatted address field in my Platlot-Address
table had to be re-joined to the parcel-map table in Arcview. This involved: The next phase was actually performing the mapping
of the different crime categories from the database. The following steps
were taken for each crime category table: This process resulted in parcel-level maps of the crime categories. The success rate of the parcel-matching was 60% on average, meaning that 60% of the records of each type of crime were matched to the Providence parcel map, and could consequently be mapped. Although this percentage seems quite small, it was actually much larger than originally anticipated. (See 'Parcel Mapping Dilemma') 4. Manipulating the Resulting Crime Maps and Creating Crime Density Maps To examine the new crime shapefiles and begin looking
at the crime trends around my case study parks, I opened the following
themes in a new view:
As I had created almost 40 themes for the various crime categories of interest, it was immediately apparent that I would have difficulty interpreting such an abundant amount of information if I didn't perform some other manipulations. I noticed that several of the crime categories I had mapped actually contained very few violations. Going back to the original Excel tables, I performed counts of the number of violations in each crime type. This made it clear that some of the crime categories had so few violations that they would lend little to a comprehensive analysis. I decided that I would remove all those types that had fewer than 25 violations, unless they were a particularly serious or interesting type of crime, such as homicide. My advisor Harold Ward then suggested that I group
similar types of crime together; i.e. assembling the different types of
violent crime. Consequently, I constructed the following groupings: In Arcview, I connected the dbf tables of the crime categories in each group. For example, I joined the tables of narcotics violations and narcotics investigations into a single table entitled 'Drug Violations'. Then, following the same basic mapping steps that were performed for each crime type (see 'Parcel mapping the crime database') I converted these new crime group tables into their own shapefiles. The resulting maps looked like the following:
To further aid my interpretation of the mapped crime data, I decided to perform density calculations on the maps. The density calculation tool in Arcview allows one to translate data into a grid format that takes into account the distribution and concentration of the data. In effect, it produces a 'hot spot' map, showing where the data points are located, and how densely they converge and/or overlap. I wanted to produce density maps that reflected the information as accurately as possible. I knew that if I used the parcel-level maps in the density calculations, I would only see a part of the picture - recall that due to data constraints, I only achieved a 60% mapping success rate. Because the density maps would not require the fine resolution that parcel-mapping gives, I decided to geocode the data before performing the calculations. (Geocoding is a process where each piece of information associated with an address is mapped as a point on a line representing a street. It allows more information to be mapped, because instead of requiring a specific parcel to be associated with each address, it recognizes an address name, extrapolates as to where that address would be on a street, and maps it as a point in the approximate location. For example, if 42 John Street wasn't in the address database, whereas 38 John Street and 50 John Street were, then Arcview would put a dot in between the two addresses it already recognized.) City-wide Crime Density Maps I made the newly drawn circle for each park it's
own shapefile, and then proceeded to 'clip' each geocoded crime-group
theme so that I would include only the crime data for that circle region.
Next, I performed density calculations on the themes to achieve density
maps for each of the crime groups. Lastly, I integrated all the crime
themes using the Map Calculation tool to get a total crime density map
for each park.
|
||||||||||||
