Support US Census Reverse Geocoding

[jessecambon]

The US Census service supports reverse geocoding through searchtype='coordinates'. See code example and data output format here.

We could potentially support this in tidygeocoder, but the output format is quite different from other geocoding services so it would require some data formatting to work properly within the package.

[LizBlackman]

Hi Jesse,
Is this enhancement going to go live anytime soon?

[jessecambon]

Hi @LizBlackman, unfortunately I don't think I'll have availability in the near future so probably not. However, you are welcome to install the branch I was working with and try it out:

devtools::install_github("jessecambon/tidygeocoder", ref = "census-reverse-geocoding")

You should then be able to use method="census" in reverse_geo() or reverse_geocode(). I was conflicted on how or if to implement this capability, because the Census reverse geocoder provides data in a very different format than the other supported services (multiple rows for each coordinate where each row is a different type of geography). You can see an example with a linked script here: #165 (comment)

Alternatively, you could use the script referred to in the link above to reverse geocode with the census service directly (ie. not using tidygeocoder). Hope that helps and feel free to give some feedback on output data format/etc.

[LizBlackman]

Hi @jessecambon!

Thank you for the quick response!

I worked on this for quite some time on Friday and went through stack overflow to find a way to clean up the output that I got back from the reversegeo() function. i needed to manually geocode a ~3200 lat/long combos. I did this thorugh a loop and got quite a few lists. From this large list i had to simplify that data to get teh CTs that i needed. While it is not clean, i got what i needed from the output. See my code below:
`GEOID1 <- apply(lat_long1, 1, function(row) reverse_geo(row['Latitude'], row['Longitude'], method="census", full_results=TRUE, custom_query = list(benchmark= "Public_AR_Census2020", vintage="Census2010_Census2020")))

alltypes<- unique(unlist(sapply(c(1:length(GEOID1)), function(x){
unlist(lapply(GEOID1[[x]]$geography, function(l) l[[1]]))
})))

all.values <- lapply(c(1: length(GEOID1)), function(x){
types <- unlist(lapply(GEOID1[[x]]$geography, function(l) l[[1]]))
matches <- match(alltypes, types)
values <- GEOID1[[x]]$geoid[matches]
})

all.values <- do.call("rbind", all.values)
all.values <- as.data.frame(all.values)
names(all.values) <- alltypes

lat_long1[, alltypes]<-NA
lat_long1[,alltypes]<- all.values
lat_long1 <-lat_long1 %>%
rename(GEOID='Census Tracts') %>%
select(TempID:census_block,
GEOID)`

Assistance came from this stack overflow entry

[jessecambon]

Awesome, glad you got it working. I'm assuming you may not be able to share these coordinates you are using, but could you possibly provide an end-to-end example that I could run with a few sample coordinates that shows the end data format? You could use something like reprex or whatever is easiest.

Also, just to confirm - you are using the census-reverse-geocoding branch version of tidygeocoder, correct?

[LizBlackman]

Hi Jesse,

I learned how to use reprex just for you. Let me know if what i shared works, it should...

test_reverse.csv

lat_long1<- read.csv("test_reverse.csv")

devtools::install_github("jessecambon/tidygeocoder", ref = "census-reverse-geocoding")
library(tidygeocoder)




GEOID1 <- apply(lat_long1, 1, function(row) reverse_geo(row['Latitude'], row['Longitude'],  method="census", full_results=TRUE, custom_query = list(benchmark= "Public_AR_Census2020", vintage="Census2010_Census2020")))

alltypes<- unique(unlist(sapply(c(1:length(GEOID1)), function(x){
  unlist(lapply(GEOID1[[x]]$geography, function(l) l[[1]]))
})))

all.values <- lapply(c(1: length(GEOID1)), function(x){
  types <- unlist(lapply(GEOID1[[x]]$geography, function(l) l[[1]]))
  matches <- match(alltypes, types)
  values <- GEOID1[[x]]$geoid[matches]
})

all.values <- do.call("rbind", all.values)
all.values <- as.data.frame(all.values)
names(all.values) <- alltypes

lat_long1[, alltypes]<-NA
lat_long1[,alltypes]<- all.values

view(lat_long1)