Skip to content

Latest commit

 

History

History
243 lines (196 loc) · 6.62 KB

hdi_tv.md

File metadata and controls

243 lines (196 loc) · 6.62 KB
Raw

Regression and Other Stories: Human Development Index

Andrew Gelman, Jennifer Hill, Aki Vehtari 2021-04-20

Tidyverse version by Bill Behrman.

Human Development Index - Looking at data in different ways. See Chapter 2 in Regression and Other Stories.

# Packages
library(tidyverse)
library(haven)
library(sf)

# Parameters
  # Human Development Index data
file_hdi <- here::here("HDI/data/hdi.dat")
  # Data that includes state income
file_income <- here::here("HDI/data/state vote and income, 68-00.dta")
  # Common code
file_common <- here::here("_common.R")

#===============================================================================

# Run common code
source(file_common)

2 Data and measurement

2.1 Examining where data come from

Data

Human Development Index data.

hdi <- 
  file_hdi %>% 
  read.table(header = TRUE) %>% 
  as_tibble() %>% 
  select(state, hdi, hdi_rank = rank, canada_dist = canada.dist)

hdi
#> # A tibble: 51 x 4
#>    state              hdi hdi_rank canada_dist
#>    <chr>            <dbl>    <int>       <int>
#>  1 Connecticut      0.962        1           2
#>  2 Massachusetts    0.961        2           2
#>  3 New Jersey       0.961        3           2
#>  4 Washington, D.C. 0.96         4           4
#>  5 Maryland         0.96         5           3
#>  6 Hawaii           0.959        6           2
#>  7 New York         0.959        7           1
#>  8 New Hampshire    0.958        8           1
#>  9 Minnesota        0.958        9           1
#> 10 Rhode Island     0.958       10           3
#> # … with 41 more rows

State income data for 2000.

income <- 
  file_income %>% 
  read_dta() %>% 
  filter(st_year == 2000) %>% 
  select(
    state = st_state,
    state_abbr = st_stateabb,
    income_2000 = st_income
  ) %>% 
  mutate(income_2000_rank = min_rank(-income_2000))

income
#> # A tibble: 50 x 4
#>    state       state_abbr income_2000 income_2000_rank
#>    <chr>       <chr>            <dbl>            <int>
#>  1 Alabama     AL              21656.               44
#>  2 Alaska      AK              27210.               15
#>  3 Arizona     AZ              23381.               37
#>  4 Arkansas    AR              19978.               48
#>  5 California  CA              29581.                8
#>  6 Colorado    CO              30406.                7
#>  7 Connecticut CT              37808.                1
#>  8 Delaware    DE              28128.               13
#>  9 Florida     FL              25978.               20
#> 10 Georgia     GA              25502.               26
#> # … with 40 more rows

hdi has 51 rows and income has 50. Let’s verify that the difference is the District of Columbia and that we can otherwise join on the states.

all(income$state %in% hdi$state)
#> [1] TRUE

setdiff(hdi$state, income$state)
#> [1] "Washington, D.C."

hdi_income <- 
  hdi %>% 
  full_join(income, by = "state")

hdi_income
#> # A tibble: 51 x 7
#>    state        hdi hdi_rank canada_dist state_abbr income_2000 income_2000_rank
#>    <chr>      <dbl>    <int>       <int> <chr>            <dbl>            <int>
#>  1 Connectic… 0.962        1           2 CT              37808.                1
#>  2 Massachus… 0.961        2           2 MA              34401.                3
#>  3 New Jersey 0.961        3           2 NJ              34963.                2
#>  4 Washingto… 0.96         4           4 <NA>               NA                NA
#>  5 Maryland   0.96         5           3 MD              31213.                5
#>  6 Hawaii     0.959        6           2 HI              25892.               22
#>  7 New York   0.959        7           1 NY              31799.                4
#>  8 New Hamps… 0.958        8           1 NH              30431.                6
#>  9 Minnesota  0.958        9           1 MN              29173.               10
#> 10 Rhode Isl… 0.958       10           3 RI              26620.               18
#> # … with 41 more rows

Human Development Index vs. average state income.

hdi_income %>% 
  drop_na(income_2000) %>% 
  ggplot(aes(income_2000, hdi)) +
  geom_point() +
  ggrepel::geom_text_repel(aes(label = state_abbr)) +
  labs(
    title = "Human Development Index vs. average state income",
    x = "Average state income in 2000",
    y = "Human Development Index"
  )

Human Development Index rank vs. state income rank.

hdi_income %>% 
  drop_na(income_2000_rank) %>%
  ggplot(aes(income_2000_rank, hdi_rank)) +
  geom_point() +
  ggrepel::geom_text_repel(aes(label = state_abbr)) +
  scale_x_reverse() +
  scale_y_reverse() +
  labs(
    title = "Human Development Index rank vs. state income rank",
    x = "State income rank in 2000",
    y = "Human Development Index rank"
  )

Boundaries for U.S. states using ussf package.

us <- ussf::boundaries(geography = "state")
setdiff(us$NAME, hdi$state)
#> [1] "District of Columbia"

setdiff(hdi$state, us$NAME)
#> [1] "Washington, D.C."

Join boundaries to Human Development Index data.

us_hdi <- 
  us %>% 
  left_join(
    hdi %>% 
      mutate(
        state = str_replace(state, "Washington, D.C.", "District of Columbia")
      ),
    by = c("NAME" = "state")
  )

Human Development Index by state.

us_hdi %>% 
  ggplot() +
  geom_sf(aes(fill = hdi)) +
  scale_fill_distiller(palette = "YlGnBu", direction = 1) +
  theme_void() +
  theme(plot.title = element_text(hjust = 0.5)) +
  labs(
    title = "Human Development Index by state",
    fill = "HDI"
  )

Number of states you need to drive through to reach the Canadian border.

us_hdi %>% 
  ggplot() +
  geom_sf(aes(fill = factor(canada_dist))) +
  scale_fill_brewer(palette = "YlGnBu", direction = -1) +
  theme_void() +
  theme(plot.title = element_text(hjust = 0.5)) +
  labs(
    title = 
      "Number of states you need to drive through to reach the Canadian border",
    fill = NULL
  )