Create Indicators for Proximity to Highway
We want to account for the effect of being very close to a busy highway on property values. The preliminary work on this has already been done by Mr. Robert Denne:
From: Robert Denne rcdenne@gmail.com Sent: Thursday, June 20, 2019 1:09 PM To: Robert Ross Subject: RE: data granularity enhancement project; First success Attachments: NumericPinsOfMajorRdBufferedParcels.csv; StepwiseModelCoefficientsFinal20190620.pdf I finally have a success to report. I’ve found several road-type sources, but the one I found most useful is at the following URL: https://hubcookcountyil. opendata.arcgis.com/datasets/4569d77e6d004c0ea5fada54640189cf_5 There a dataset is available that includes “Major Road” and also a (so-far useless) feature “Type of Road”. The latter includes codes beyond the seven defined in other road datasets I’ve seen, such as those at the next URL: http://www.idot.illinois.gov/Assets/uploads/files/Transportation-System/Maps-&-Charts/Five- Year/FCCounty/FC%20County_Cook.pdf The seven-valued set I believe is based on federal funding for certain roads according to their so called functional classification. I think the seven would probably be the most useful down the road, but since “Major road” was unambiguous and readily available I chose it for my proof of concept test. By using a GIS I was able to create a 100 foot buffer around the “Major roads” from the dataset downloaded from the first site above, which I then plotted with the parcel dataset from you, limiting the analysis to five Northern-Tri townships: Palatine, Wheeling, Northbrook, Elk Grove, and Maine. Parcels that intersected the buffer were extracted, along with their PINs, which allowed me to flag the relevant parcels in the modelling dataset that I also have from you. I then created a (kitchen-sink) stepwise model for the five townships and obtained a coefficient indicating that single family residential parcels next to major roads have a value about 12 percent less than they would otherwise have. I’m attaching the coefficients from the last SPSS stepwise output. Your modeling efforts are undoubtedly superior to mine, but I wanted to establish that the new feature was worth incorporating before refining the structure of the model. I’m also attaching a list of parcel PINs that would allow you to replicate & enhance my work if you’re interested in doing so. Unfortunately I see that most of the relevant townships in my sample have already been done. If you want I could walk you through the process I followed in some greater detail, but at the moment it’s not yet as well organized as I’d like, due to GIS problems and backtracking faced along the way. I’d be interested to hear what you think about this, and any other data enhancements you think it productive to pursue. Cheers, Bob
StepwiseModelCoefficientsFinal20190620.pdf
Please alter 0_build_pin_geography.R to
-
Ingest this road shape file. -
Create a series of buffers for the interstate highways 50ft(50ft)300ft -
Using the current PIN shape file, identify which PINs are within each buffer.
At this point, you should have a data set that has 5 new columns which are binary indicators for proximity to a roadway.
-
Please modify the appropriate data dictionary file to create 5 new models which are variants of the current 'best' model that include each of the 5 proxies. -
Execute the Residential script on all townships we have currently estimated to create 'TEST' scope files. -
Create a report in excel of the improvements in the modeling process due to the inclusion of these indicators.
@ssmithassessor @ccaojardine this project may interest you. This is a test case for #116 (closed).