Skip to main content

Table 2 Distance between locations for the same address assigned by two geocoding methods, by area characteristics

From: Comparing a single-stage geocoding method to a multi-stage geocoding method: how much and where do they disagree?

  Discrepancy-distance (m)
      Percentiles  
   N Arith. mean (95% CI) Geo. mean (95% CI) 50th 90th 95th 99th Percent > 1 km apart
Overall   7,686 160 (140, 179) 49 (48, 51) 54 180 296 2218 1.8%
Census Tract Poverty* ≥ 20% 1,027 156 (77, 234) 28 (25, 30) 26 144 258 1370 1.4%
  10 to 19% 2,248 169 (135, 203) 46 (44, 49) 48 183 310 3052 2.2%
  < 10 % 4,411 156 (132, 180) 59 (57, 61) 62 187 300 1873 1.7%
Census Tract Density* (population/km 2 ) ≥ 1000 3,770 106 (83, 129) 44 (43, 46) 50 134 174 717 0.8%
  500 – 999 1,504 141 (108, 174) 50 (47, 53) 55 175 269 1970 1.6%
  200 – 499 931 213 (140, 286) 49 (45, 54) 52 202 333 5593 2.5%
  < 200 1,481 281 (218, 345) 66 (61, 71) 70 401 779 5548 4.1%
Most common reference for multi-stage method* Local Parcels 4,644 152 (128, 176) 61 (59, 62) 63 172 258 1704 1.5%
  Local Roads 1,075 136 (107, 165) 52 (48, 56) 53 193 333 1667 2.0%
  TIGER-based 1,967 191 (140, 242) 30 (28, 32) 25 211 414 4028 2.5%
  1. Discrepancy-distance indicates distance between multi-stage and single-stage geocoded address coordinates for the same address; p50, p90, p95, and p99 indicate the 50th, 90th, 95th, and 99th percentiles; TIGER indicates Topologically Integrated Geographic Encoding and Referencing system line files, this includes NAVTEQ and Dynamap reference files
  2. * Indicates significance (p < 0.05) for a comparison of discrepancy-distances across subgroups using a linear regression model