Skip to main content

Table 2 Likelihood-based parameter estimates for the best-fitting models.

From: Modeling the probability distribution of positional errors incurred by residential address geocoding

Error dataset Component Proportion μ X μ Y σ X σ Y ρ v
(a) 1 0.571 -12.1 -10.7 61.6 54.1 -0.05 1.6
  2 0.253 -4.7 -350.0 75.9 550.0 0.18 6.5
  3 0.176 352.8 -12.6 540.3 84.9 -0.03 16.7
(b) 1 0.560 -0.8 -14.2 39.4 75.9 0.06 1.8
  2 0.440 372.1 -6.7 523.6 90.3 -0.10 5.9
(c) 1 0.519 4.9 -5.4 62.3 60.8 -0.10 1.8
  2 0.292 13.6 -35.0 289.1 54.9 -0.14 2.4
  3 0.189 14.9 -10.2 62.1 354.4 0.14 2.4
(d) 1 0.700 5.9 -4.3 47.0 100.7 0.06 1.8
  2 0.300 29.3 -6.2 62.1 419.5 0.16 3.0
  1. Models and the datasets to which they were fitted are: (a) the three-component t mixture model for the automated geocoding positional errors; (b) the two-component t mixture model for the automated geocoding positional errors aligned with axial direction of corresponding street segment; (c) the three-component t mixture model for the E911 positional errors; (d) the two-component t mixture model for the E911 positional errors aligned with axial direction of corresponding street segment. Means are denoted by μ X and μ Y , standard deviations by σ X and σ Y , correlation coefficient by ρ, and degrees of freedom by v. Units of measurement for means and standard deviations are meters.