Skip to content

Instantly share code, notes, and snippets.

@riordan
riordan / getting-there-thoughts-on-interpolation-in-pelias.md
Created February 22, 2016 17:34
A short note on how we might go about implementing interpolation in Pelias

To know everything about everywhere is a pretty substantial goal. To find every address in the United States could require having surveyed and found points for every address, and then re-surveying every time there’s an additional address.

When we search for a place that we don’t have a specific point for, we could possibly return another address on the same street (if we have a match there), there’s not much otherwise we can or should do.

Obviously, this is ideal for data quality, but far from ideal for a data management standpoint.

Interpolation is designed to fill this gap. By building on top of the road networks, which are far more available, interpolation allows forward and reverse geocoding without the need to have rooftop points for all possible places. It also ensures that new construction within the existing road networks can be found without needing to add additional points.

Importantly, interpolation can also play very well with point data, taking point data everywhere it can get it, and falling

Prerequisites: Installing R + RStudio

Brown Institute Transparency Workshop with Harry Enten & Neil Paine of 538 (9/24/16)

To participate in the workshop, you'll have to install some software on your computer prior to the workshop. It'll probably take about a half hour to get everything set up.

These instructions cover installation for Mac OS and Windows. If you're a Linux user, you have my admiration and esteem and can probably figure this out for yourself (be sure to install the plugins).

If you have any issues with this, don't hesitate to contact David Riordan, Rosalie Yu, or Michael Krisch for assistance.