diff --git a/docs/02-geocoding.Rmd b/docs/02-geocoding.Rmd old mode 100644 new mode 100755 index a25baa3..fe744a8 --- a/docs/02-geocoding.Rmd +++ b/docs/02-geocoding.Rmd @@ -416,4 +416,15 @@ all_distances = all_distances[which(!is.na(all_geocodes$truth.lat)),] save(all_distances, file = '../results/all_distances.rda') ``` +### Compute Supplementary Statistics for +Upon review, it was recognized that RMSE alone is an insufficient measure of accuracy. So in addition, we compute percentiles [median (50), 75, 90, max (100)] for distances of each geocoder. + +```{r} +load('../results/all_distances.rda') +dist_ntiles <- apply(all_distances[,2:15], 2, function(result){ + quantile(result, na.rm = TRUE, probs = c(.5, .75, .9, 1)) %>% round(1) +}) +print(dist_ntiles) +write.csv(dist_ntiles, file = '../results/ntile_distances.csv') +``` diff --git a/docs/02-geocoding.nb.html b/docs/02-geocoding.nb.html old mode 100644 new mode 100755 index 6e391c6..9581676 --- a/docs/02-geocoding.nb.html +++ b/docs/02-geocoding.nb.html @@ -1,11 +1,10 @@ - + - @@ -15,6 +14,19 @@ Test Geocoders + + +