Past records not in any assembly district

gregallensworth commented 1 year ago

Discovered while updating polygon data in #38

The crashes_all_prod table has 2,071,732 crashes in it today, of which 205,560 (about 10%) have NULL for assembly

That's about 10% which "feels" like a lot to fall outside of any assembly district but (presumably) fitting into the five boroughs.

I should check into this,

see how this many fall outside any Assembly District
and see to what degree this also means City Council, Neighborhood, Community Board, ...

┆Issue is synchronized with this Asana task

gregallensworth commented 1 year ago

Initial overview

I exported a shapefile of the 205,560 crash-points SELECT * FROM crashes_all_prod WHERE assembly IS NULL

And I exported the assembly as GeoJSON. This is the old 2015 shapes, but that fits our intent, finding why the past records have been coming up this way.

Initial impressions are that this is indeed an edge effect. But many of them seem as if they would not be solved with a simple buffer (even if CARTO would let that happen without timing out).

Many are on bridges which are not covered by these polygons. Those may require desktop GIS work to find the closest district and assign them.
Many are "simply" edge effects where a combination of bad coordinates + districts data differing from streets data add up to put a crash a few meters outside of any area. Some are 20+ meters outside of an area.
A few are well outside of any area district and those seem very odd. Maybe data entry error? Maybe a very tiny island? Maybe a bicyclist hit someone whilst on a ferry?