Screening Methods, Predictions, and Modeling
This map shows the City of Chicago and all land use polygons of the four desired types, part of Screening Method 2.
This section describes how inequity was identified in order that it could be mitigated. Includes: importance of data, how the CDOT Bicycle Parking Program manages data, how the Underserved Wards were selected to be part of the project (Screening Methods), alternative ways that underserved areas can be identified (using GIS), and an introduction to the Bike Parking Demand Prediction Model.
Before I discuss the screening methods, I introduce some basic and preliminary analysis of the expansive and well-managed bicycle parking dataset.
Analysis
Statistics
General arithmetic statistics about the Bike Parking Program bike rack installation dataset.[A]
NOTE: The data in the table below is current as of Monday, February 15, 2010. The dataset used in this paper is incomplete and only includes installation data for the following years: 1999, 2000, 2001, 2002, 2007, 2008, and 2009. The Bike Parking Program has installed bike racks in all 50 Wards in Chicago.
| Maximum | 1056 (Ward 42) |
|---|---|
| Minimum | 48 bike racks (Ward 12) |
| Mean | 176 bike racks |
| Median | 125 bike racks |
Equitable distribution factor
See main article, Equitable distribution factor.
Data Requirements
No matter which Screening Method you perform, or if you perform a custom method of identifying underserved areas, you will need good data about already existing bike parking facilities. Screening Method 2 requires additional data, including shapefiles for the Census Tracts of the service area under study, and the locations of bikeways.
For assistance on managing bike parking data (for users with an existing database, or users who want to start a new database), please see my paper Best practices for managing bicycle parking data.
Importance of Good Data
To analyze bike parking distribution you need to have a good dataset that has the locations and quantities of bike parking facilities in your locality. It's most helpful if you have ascribed the area (region) attribute on which you want to divide the records to determine your distribution - in other words, if you want to sort by ZIP code, ensure you've geocoded the location and included its ZIP code.
Data History
Prior to my employment at the Chicago Department of Transportation, the Bicycle Parking Program stored data in various Microsoft Excel or Corel Quattro spreadsheets created by different staff members. It seemed they used random layouts and structures. A Microsoft Access database superseded the spreadsheets, but all data saved prior to the Access database remained unconverted to the new format. In 2007, I developed a new database using MySQL[a] and accessed via a custom web application I built in PHP[b].
Current Data
As of January 17, 2010, the BPP database has 5,234 records with installation information for 8,799 bike racks (of all types). The majority of these records are “historical;” they existed within the various files and formats prior to the creation of the MySQL database. The database represents an amalgamation of data from all discovered sources. It would be ideal to have included records of all 12,145 installed bike racks[B], [C].
Data Shortcomings
See more Project shortcomings.
The Bike Parking Program database lacks installation data for 1993-1998, and 2003-2006. It has data indicating the total number of bike racks installed per year, but not their locations. Because of this, we cannot fully determine the extent of the story about inequitable distribution. It is quite possible that, in the missing data, there's information that shows equity was achieved for certain years.
Good Data Management
After upgrading the Bike Parking Program data management system, I wrote a white paper for wide publication to educate other Bike Parking programs around the world on how to setup a similar database and avoid the same problems.
See main article, Good data management.
Screening Methods
I developed two “screening methods” to help analyze the data. A screening method, in this context, is a set of tasks to make preliminary and targeted identifications of desired attributes within a dataset.
The first screening method compares each ward to every other ward using the most current data. It is simple to perform this analysis and has a single variable. The result is a list of Wards that have disproportionately fewer bike parking facilities.
The second screening method applies geographic criteria on multiple spatial datasets to find specific parcels to survey for potential bike parking facilities.
Screening Method 1
See main article for Screening Method 1.
This was the actual method I used to “discover” the 13 Underserved Wards. Requires the following tools and skills:
- Spreadsheet software
- Functions within spreadsheet software
- Knowledge of mathematic averages
What does Screening Method 1 look like now, in 2010? Did the status of any of the 13 Underserved Wards change?
Screening Method 2
See main article for Screening Method 2.
Using Geographic Information System (GIS)[c] software to analyze land use.
- GIS software. ArcGIS from ESRI is very common, although free and open-source options are available . I don't know if they have the essential functions to perform the analysis in Screening Method 2.
- Knowledge of GIS software
- Additional datasets, including the location of bike racks
The analysis in Screening Method 1 is very simple. While GIS software and the necessary data were available to me, I had neither the knowledge nor skills to develop within a suitable model or process that would accomplish analysis equal or similar to that I discuss in Screening Method 2. Finally, this project includes a theoretical demand prediction model that, given a list of addresses, can help prioritize locations for potential bike parking facilities.
Bike Parking Demand Model
Main article: Bike Parking Demand Model.
Land Use Analysis
Main article: Land use analysis.
