Match | Mean | Distance | Misclassification Rates | Census Tract |
|---|
Type
|
ρ
|
< 100 m
|
False +
|
False –
|
Total
|
Concordance
|
|---|
Street | 0 | 27% | 0% | 0% | 0% | 100% |
150 | 29% | 8% | 6% | 15% | 90% |
300 | 26% | 11% | 11% | 22% | 82% |
600 | 27% | 15% | 14% | 29% | 70% |
Centroid | 0 | 32% | 0% | 0% | 0% | 100% |
2,500 | 19% | 9% | 22% | 31% | 66% |
5,000 | 16% | 9% | 25% | 33% | 55% |
10,000 | 14% | 8% | 26% | 34% | 42% |
- aSpherical distance in meters between criterion standard and vendor-assigned coordinates. Standard deviation of ρ = 500 and 15,000 meters for street- and centroid-type matches, respectively. bInterstate, U.S., or state highway or major traffic thoroughfare. cFalse + indicates misclassification of the unexposed (≥ 100 m) as exposed (< 100 m). False – indicates misclassification of the exposed as unexposed. The sum of false + and – error rates may not equal the total error rate due to rounding. dPercent of census tracts matching those in the datasets without positional error (ρ = 0). Based on a 5% random sample of street-type address matches (n = 2,608) and a census of centroid-type address matches (n = 2,671) in The Environmental Epidemiology of Arrhythmogenesis in WHI, 1999–2002.
