datakind / new-america-housing-loss-public

MIT License
10 stars 5 forks source link

Release #32

Closed kaylawilding closed 1 year ago

kaylawilding commented 1 year ago

DKW-631: Output list of exceptions for uncoded addresses DKW-629: Improve handling of FEAT's missing directories DKW-630: Eviction summary counts vary slightly between runs, make constant if using same input files I did a lot of digging and testing and found some helpful info on the FAQs of the census geocoder site. They stated that inconsistency in their results is a known issue and they are working to fix it. (https://www2.census.gov/geo/pdfs/maps-data/data/Census_Geocoder_FAQ.pdf at pg4-5) Their suggestion was to re-run those that don't match through the geocoder again. It now tries to geocode three times (I did some testing and it seems across three runs, we should hit all of the addresses that will hit a match) to get at those that incorrectly return a no match. It only geocodes those that don't match, thus it doesn't seem to add a substantial amount of time when running. In the geocoded data, it now is blank rather than saying "No Match" or "Tie" in the match column. DKW-636 : Resolve cases of dropped addresses DKW-1036: Resolve dropped records issue in NYC data [Multi-county issue] DKW-1045 : Investigate why the # of census tracts are off [This is the Zip coding issue, so removing the zip code geocoding]