Abstract:
Fundamentos: Geocodificar es asignar coordenadas geográficas a
puntos del espacio, frecuentemente direcciones postales. El error cometido
al aplicar este proceso puede introducir un sesgo en las estimaciones
de modelos espacio-temporales en estudios epidemiológicos. No se
han encontrando estudios q... Ver más
spatial points, which often are postal addresses. The error made in applying
this process can introduce bias in estimates of spatiotemporal models in epidemiological
studies. No studies have been found to measure the error made in
applying this process in Spanish cities. The objective is to evaluate the errors
in magnitude and direction from two free sources (Google and Yahoo) with
regard to a GPS in two Spanish cities.
Method: 30 addresses were geocoded with those two sources and the
GPS in Santa Pola (Alicante) andAlicante city. The distances were calculated
in metres (median, CI95%) between the sources and the GPS, globally
and according to the status reported by each source. The directionality
of the error was evaluated by calculating the location quadrant and
applying a Chi-Square test. The GPS error was evaluated by geocoding 11
addresses twice at 4 days interval.
Results: The overall median in Google-GPS was 23,2 metres (16,0-32,1)
for Santa Pola, and 21,4meters (14,9-31,1) forAlicante. The overallmedian in
Yahoo was 136,0 meters (19,2-318,5) for Santa Pola, and 23,8 meters (13,6-
29,2) for Alicante. Between the 73% and 90% were geocoded by status as
“exact or interpolated” (minor error), where Goggle andYahoo had a median
error between 19 and 23 metres in the two cities. The GPS had a median error
of 13.8 meters (6,7-17,8). No error directionality was detected.
Conclusions: Google error is acceptable and stable in the two cities, so
that it is a reliable source for geocoding addresses in Spain in epidemiological
studies.
|