###### 1
# Load up the R packages we need for the Census data

install.packages("tidyverse")
install.packages("tidycensus")

library(tidycensus)
library(tidyverse)

census_api_key("cc2610e9d3d6c8c926494884bcbe30a647038dac")

###### 2
# Let's get the population of each state with the Census API for the
# year 2010.

pop <- get_decennial(geography = "state",
                     variables = "P001001",
                     year = 2010)

glimpse(pop)

###### 3
# We can get more than one attribute at once, this is the 
# population from before, and the population identified in 
# rural areas.

rural <- get_decennial(geography = "state",
                       variables = c("P001001", "P002005"),
                       year = 2010,
                       output = "wide")
glimpse(rural)

###### 4
# Now, let's plot this as a scatter plot. First, we add a new 
# column for the fraction of population that is in rural areas,
# then reorder the states by this fraction, and finally add in 
# the labels for the graph.

rural %>%
  mutate(prop_rural = P002005/P001001) %>% 
  ggplot(aes(x = prop_rural, y = fct_reorder(NAME, prop_rural))) +
  geom_point() +
  labs(title = "Rural Population in US States in 2010",
       subtitle = "Maine and Vermont are the most rural states",
       caption = "Source: US Census",
       x = "Rural Population Proportion",
       y = NULL)

###### 5
# The new piece we want to add today is the geometry of each state
# (or whichever locale we query from the Census). You can see in 
# this table that there is a new column filled with polygons for the 
# shapes of the states. These are denoted with GPS coordinates.

rural <- get_decennial(geography = "state",
                       variables = c("P001001", "P002005"),
                       year = 2010,
                       output = "wide",
                       geometry = TRUE) 
glimpse(rural)

###### 6
# Using a new plot type, the geom_sf(), we can draw these shapes
# and fill them in based on the proportion of rural population.

rural %>%
  mutate(prop_rural = P002005/P001001) %>% 
  ggplot(aes(fill = 100 * prop_rural)) +
  geom_sf() + 
  scale_fill_viridis_c(option = "plasma",
                       direction = -1) +
  labs(title = "Rural geography of the United States",
       caption = "Source: Census 2010",
       fill = "Percent Rural") +
  theme_void()

###### 7
# The above plot is hard to see because of the distant states and Puerto Rico.
# We can filter them out by adding in the second line.

rural %>%
  mutate(prop_rural = P002005/P001001) %>% 
  filter(! NAME %in% c("Alaska", "Hawaii", "Puerto Rico")) %>%
  ggplot(aes(fill = 100 * prop_rural)) +
  geom_sf() + 
  scale_fill_viridis_c(option = "plasma",
                       direction = -1) +
  labs(title = "Rural geography of the United States",
       caption = "Source: Census 2010",
       fill = "Percent Rural") +
  theme_void()

###### 8 
# Or we can use a different map projection, and shift Alaska and Hawaii
# into a better location for graphing

rural_shifted <- get_decennial(geography = "state",
                               variables = c("P001001", "P002005"),
                               year = 2010,
                               output = "wide",
                               geometry = TRUE,
                               shift_geo = TRUE) %>%
  rename(state = NAME) %>%
  mutate(prop_rural = P002005/P001001)

rural_shifted %>%
  ggplot(aes(fill = prop_rural * 100)) +
  geom_sf() + 
  scale_fill_viridis_c(option = "plasma",
                       direction = -1) +
  labs(title = "Rural geography of the United States",
       caption = "Source: Census 2010",
       fill = "Percent Rural") +
  theme_void()

###### 9
# Let's dive in to one state, West Virginia, and look at some data 
# at the county level. First is the median income.

wv <- get_acs(geography = "county", 
              variables = c(medincome = "B19013_001"), 
              state = "WV", 
              year = 2018,
              geometry = TRUE)
wv

wv %>%
  ggplot(aes(fill = estimate)) +
  geom_sf() + 
  scale_fill_viridis_c(option = "plasma",
                       direction = -1) +
  labs(title = "Household income by county in West Virginia",
       subtitle = "2014-2018 American Community Survey",
       fill = "Household income") +
  theme_void()

###### 9
# And second is the rural population proportion. We can note the correlation
# between these two maps.

rural_wv <- get_decennial(geography = "county",
                       variables = c("P001001", "P002005"),
                       year = 2010,
                       output = "wide",
                       state = "WV",
                       geometry = TRUE) 

rural_wv %>%
  mutate(prop_rural = P002005/P001001) %>% 
  ggplot(aes(fill = 100 * prop_rural)) +
  geom_sf() + 
  scale_fill_viridis_c(option = "plasma",
                       direction = -1) +
  labs(title = "Rural geography of West Virginia",
       caption = "Source: Census 2010",
       fill = "Percent Rural") +
  theme_void()

###### 10
# Finally, we can see how to split up and draw multiple maps
# with a facet wrapping. Here we look at the racial makeup of
# Harris county in Texas, the home of Houston.

racevars <- c(White = "B02001_002", 
              Black = "B02001_003", 
              Asian = "B02001_005",
              Hispanic = "B03003_003")
harris <- get_acs(geography = "tract",
                  variables = racevars, 
                  year = 2018,
                  state = "TX",
                  county = "Harris County",
                  geometry = TRUE,
                  summary_var = "B02001_001") 
harris %>%
  mutate(Percent = 100 * (estimate / summary_est)) %>%
  ggplot(aes(fill = Percent, color = Percent)) +
  facet_wrap(~ variable) +
  geom_sf() +
  scale_fill_viridis_c(direction = -1) +
  scale_color_viridis_c(direction = -1) +
  labs(title = "Racial geography of Harris County, Texas",
       caption = "Source: American Community Survey 2014-2018") +
  theme_void()

###### 11
# Your turn! 

# First, go back and edit these plots and queries to personalize
# the graphs to your home state and county.

# Second, search the Census for other interesting attributes that you
# would like to graph, and edit the queries to use them.