18-geocoding.Rmd

# Postcodes & addresses {#geocoding}

A common use case in our work is to do some spatial analysis on a dataset with addresses. For example, produce a map of certain incidents over the last year, or summarise the number of incidents by Ward.

## UPRN

*TODO*

## Postcodes

If there's no UPRN, getting a location via the postcode is usually the easiest option. It's not as accurate as a UPRN or a full address, but often it's accurate enough for what's required.

Below is an example of using our `add_pcd_vars()` function that builds on the [PostcodesioR](https://docs.ropensci.org/PostcodesioR/) package, which itself is a wrapper for the [postcodes.io](https://postcodes.io/) API. You pass it a data frame with a postcode column and it returns the data frame with new columns based on the postcode.

```{r add-pcd-vars, message=FALSE, warning=FALSE}
# Create a data frame with some example records
df <- tibble::tribble(
  ~name,    ~postcode,
  "SCC",    "S1 2HH",
  "Blades", "S2 4SU",
  "Owls",   "S6 1SW"
)

# # Load the function we need
# source("https://raw.githubusercontent.com/scc-pi/functions/main/Addresses.R")
# 
# # Add postcode related columns to our data frame
# add_pcd_vars(df,
#              pcd_name = "postcode",
#              .admin_district = FALSE,
#              .lat_long = TRUE,
#              other_vars = c("admin_ward",
#                             "msoa_code")) %>%
#   knitr::kable() #display them in a nice table
```

`admin_district` can be used to answer the questions: Is the postcode invalid? Is the postcode in Sheffield? You can anticipate values of `NA` and `Sheffield` respectively if the answers are yes. These are common questions so the function has an `.admin_district` argument.

Needing the coordinates of the postcode centroid is also common, so the function also has a `.lat_long` argument that if `TRUE` returns both a latitude column and a longitude column.

*TODO: link to sf example & area stats*

Other columns are added by including their name in a character vector passed with the `other_vars` function argument. For example, `other_vars=c("ccg_code","lsoa_code")`. The list of valid `other_vars` is defined at [docs.ropensci.org/PostcodesioR/\#examples](https://docs.ropensci.org/PostcodesioR/#examples).

The `pcd_name` function argument is used to specify the name of the data frame's postcode column when it isn't `postcode`. For example, `pcd_name="pupil_pcode"`.

Sometimes, the postcode in a dataset is not included as a separate column, but is at the end of a combined address column. Depending on the format and quality of the combined address column you could extract the postcode like this:

```{r, extract-pcd, message=FALSE, warning=FALSE}
# Create a data frame with some example records
df <- tibble::tribble(
  ~name,     ~address,
  "SCC",     "Pinstone St, Sheffield City Centre, Sheffield S1 2HH",
  "Blades",  "Bramall Ln, Highfield, Sheffield S2 4SU",
  "Owls",    "76 Penistone Rd N, Sheffield S6 1SW"
)

# Extract the postcode from the address into a separate column
dplyr::mutate(df, 
              postcode = stringr::word(address, -2, -1, sep = "[:space:]")) |>
  knitr::kable()
```

### PAS postcode lookup

PAS, the Performance & Analysis Service in the People Porfolio, maintains a postcode lookup table that performs the same role as the `add_pcd_vars()` function.

*TODO: Does it also include Sheffield specific variables e.g. which of the 100 neighbourhoods is the postcode in?*

### Postcode boundaries

*TODO:* [OPEN DATA GB POSTCODE UNIT BOUNDARIES](https://longair.net/blog/2021/08/23/open-data-gb-postcode-unit-boundaries/)

### Split postcodes

*TODO*

## Addresses

Anne Tetley has provided the following guidance (on Portal):

-   [Geocoding Address Data in Portal](https://sheffieldcitycouncil.cloud.esriuk.com/portal/apps/storymaps/stories/65cd11fc65ef4c0cb89b761ebf367053)


*TODO*\
- Portal LLPG Cascade Locator\
- OS Places ESRI integration