In a previous post, we had discussed investigating crime data around the 606 (Bloomingdale) trail in Chicago. One of the main issues I faced was how to identify crimes that occurred “around” the 606. Each detailed crime record contains information including:
- Ward number
- An approximate address
- Approximate Latitude
- Approximate Longitude
Using the ward number wasn’t really a possibility for a number of reasons:
- Many records didn’t have a ward on the record (can’t I just stop here ?)
- Chicago wards are irregularly shaped; thus there is no particular ward
which is always near the trail
- Wards are often redistricted , whereby the boundaries of a ward are changed.
It wasn’t clear to me whether a ward listed on the record was a snapshot at a time
or a reflection of the current ward definitions (I would guess the former)
I also considered using the approximate address. This would require maintaining a list of all streets adjacent to the trail and selecting based on a range of street addresses on this street. That’s probably doable, but prone to error and omission (mostly mine, not the crime data).
In the end I decided to define a bounding box that encompassed the entire trail. Then I could use the latitude and longitude from a record to determine whether it fell within the
bounding box and thus would be included in the analysis. Additionally, I could fairly easily alter the size of this bounding box to include a smaller or larger area to be analyzed.
Defining the bounding box
The 606 is referred to as a linear park. In reality, of course, it isn’t perfectly straight,
and unlike a line it actually has some width to it (I’d guess around 50 feet in most areas). But I think for our purposes it is close enough to a line that we can define a rectangle around it that encompasses the entire trail and the line is fairly centered in the box. Also fortunately, the line is oriented in a fairly E-W direction which our task a bit easier.
The question now becomes how big that rectangle should be. My first pass was to make it a city block in each direction in a N-S direction from the line itself. Subsequent analyses may widen or narrow this rectangle as we investigate further.
Though we won’t actually use this visual, this is what the bounding box within a city block of the trail would look like. We would use the boundaries of this box and select anything that fell within the box for an analysis. Recall this is just a first pass; one could argue that the box should be wider or extended on its ends.
Here is another view of the bounding box with the trail placed inside in green