Skip to contents

The evidence weight for confirming two records share a location at each granularity, used by the hierarchical geo score. The score takes the maximum applicable term ("how precisely is the location confirmed"), so correlated fields are never double-counted and a redundant noisy field (e.g. a wrong city when the ZIP already matches) cannot lower the score.

Usage

np_default_geo_weights()

Value

A named numeric vector.

Details

Ordered strongest to weakest: full ZIP+4 (zip9), same street and suite (unit), same street (street = number + body), 5-digit ZIP, matching PO box, 3-digit ZIP (SCF region), city, state.

unit sits just above street and below zip9: a confirmed suite in a multi-tenant building is a more precise location than the street alone, but it is not claimed to beat a full ZIP+4. It is credited only where the suite is informative — see np_addr_freq() — and, because the terms combine with max(), never subtracts when a suite is absent, differing, or shared by a crowd. A weight vector without unit (one saved before this existed) falls back to 0.97.