---
title: "Homicides in Medellin, Colombia, 2010--2019"
author: "Joe Bloggs"
date: "`r format(Sys.Date(), '%e %B %Y')`"
format:
  html:
    embed-resources: true
---

```{r setup}
#| include: false

# The R code in this document is split up into different code chunks. Each chunk
# corresponds to one of the sections of the R script files we have created
# previously, e.g. one section to load the data, another to wrangle it, etc.
# Quarto files might need more chunks than there are sections in an R script
# file, because you will probably want to add text between different maps,
# tables, etc. in the output.
# 
# Each code chunk in this document has a label at the start. This label will be
# used in any error messages produced when you render the document, making it
# easier to find the source of errors.

pacman::p_load(sf, tidyverse)
```

```{r load-data}
#| cache: true
#| include: false

# Load homicide data
homicides <- read_csv2("https://mpjashby.github.io/crimemappingdata/medellin_homicides.csv") |>
  janitor::clean_names() |> 
  remove_missing(vars = c("longitud", "latitud"), na.rm = TRUE) |> 
  st_as_sf(coords = c("longitud", "latitud"), crs = "EPSG:4326") |> 
  st_transform("EPSG:3115")

# Neighbourhoods
medellin_comunas <- read_sf("https://mpjashby.github.io/crimemappingdata/medellin_comunas.gpkg") |> 
  janitor::clean_names() |> 
  st_transform("EPSG:3115") |> 
  select(nombre, geom) |> 
  mutate(nombre = str_to_title(nombre))
```

```{r wrangle-data}
#| include: false

# Count number of homicides in each neighbourhood
homicide_count <- homicides |> 
  mutate(year = lubridate::year(fecha_hecho)) |> 
  st_join(medellin_comunas) |> 
  st_drop_geometry() |> 
  count(nombre, year)

# Convert that data into a format that is suitable for displaying in a table
homicide_table <- homicide_count |> 
  # Calculate the mean number of homicides in each neighbourhood per year
  group_by(nombre) |> 
  mutate(mean_count = mean(n)) |> 
  # Arrange the neighbourhoods by mean count of homicides in descending order
  arrange(desc(mean_count)) |> 
  ungroup() |> 
  # Convert the data to wide format for easier display
  pivot_wider(names_from = year, values_from = n, values_fill = 0)

# The next piece of code uses the `pull()` function from the dplyr package to
# extract columns from datasets (e.g. so they can be summarised using functions
# such as `max()`) and the `pluck()` function from the purrr package to extract
# single values from datasets. It also uses functions from the scales package to
# format values so that they are easier to read.
```

@tbl-homicide-counts shows the number of robberies in each neighbourhood of Medellin in each year from `r min(pull(homicide_count, "year"))` to `r max(pull(homicide_count, "year"))`. During that time, on average the `r pluck(homicide_table, "nombre", 1)` neighbourhood had the highest number of homicides (`r scales::comma(pluck(homicide_table, "mean_count", 1), accuracy = 1)` per year).

```{r neighbourhood-table}
#| label: tbl-homicide-counts
#| tbl-cap: "Robberies in neighbourhoods of Medellin"

homicide_table |> 
  # Remove columns we don't want to include in the table
  select(-mean_count) |> 
  # Remove count of offences with co-ordinates outside the city boundary
  filter(!is.na(nombre)) |> 
  # Format the table for nicer display
  gt::gt()
```