pummeler icon indicating copy to clipboard operation
pummeler copied to clipboard
verified publisher icon djsutherland

Featurization issues

Open djsutherland opened this issue 9 years ago • 1 comments

  • MIGPUMA has joint meaning with MIGSP; same for POWPUMA/POWSP.
  • Why does RELP come up so much in the ridge models? What does it mean in practice?

djsutherland avatar Jul 29 '16 05:07 djsutherland

  • CITWP, YOEP, JWMNP: mean-coding blanks might not be the right thing, since blank means the person was born in the US / doesn't work
  • MLP* (when served in military) could probably be simplified; VPS does that
  • NWAB, NWAV, NWLA, NWLK, NWRE are recoded into ESR
  • RELP (relationship to reference person) is kind of a weird feature
  • hierarchical featurization for ANC_P / FOD_P / INDP / NAICSP / OCCP / SOCP?
    • for NAICS/SOC see https://www.census.gov/people/io/methodology/indexes.html
  • merge ANC1P/ANC2P, RAC1P/RAC2P/...?
  • re-featurize JWAP/JWDP to be circular?
  • Things that refer to specific in-US places: MIGPUMA, MIGSP, POBP, POWPUMA, POWSP
  • POVPIP: pretty sharp featurization difference between 500 and 501

djsutherland avatar Nov 07 '16 17:11 djsutherland