R Econ Visual Library

# Install and load packages ---------------
packages <- c(
  "tidyverse",
  "haven",
  "latex2exp",
  "gridExtra",
  "lemon",
  "sciplot"
)

# Change to install = TRUE to install the required packages
pacman::p_load(packages, character.only = TRUE, install = FALSE)

# Load an example dataset ---------------
data <- read_dta("https://github.com/worldbank/r-econ-visual-library/raw/master/Library/Data/ScatterCI.dta")

set.seed(251668)

subdata <- data %>%
  filter(
    cons_pae_m_sine < quantile(.$cons_pae_m_sine, 0.99) & 
      cons_pae_m_sine > quantile(.$cons_pae_m_sine, 0.01)
    ) %>%
  sample_frac(0.05) # To emphasize the graded CI, 5% of obesrvations are randomly sampled

ggplot(subdata, aes(x = cons_pae_m_sine, y = cons_pae_sd_sine)) +
  geom_point(color = "black", alpha = 0.2, size = 0.8) +
  stat_smooth(aes(size = NA, fill = "90%"), alpha = 0.3, show.legend = TRUE, level = 0.90) +
  stat_smooth(aes(size = NA, fill = "95%"), alpha = 0.2, show.legend = TRUE, level = 0.95) +
  stat_smooth(aes(size = NA, fill = "99%"), alpha = 0.1, show.legend = TRUE, level = 0.99) +
  stat_smooth(color = "black", size = 0.8, alpha = 0.6, se = FALSE, show.legend = TRUE) +
  theme_classic() +
  scale_fill_manual(
    name = "Confidence Intervals", 
    values = c("90%" = "gray12", "95%" = "gray12", "99%" = "gray12")
    ) +
  guides(
    fill = guide_legend(override.aes = list(
      alpha = c(0.3, 0.2, 0.1),
      size = c(0, 0, 0))
      )
    ) +
  xlim(c(75, 130)) +
  ylim(c(5, 22)) +
  xlab(TeX("$\\widehat{m}$")) +
  ylab(TeX("$\\widehat{s}$")) +
  theme(
    axis.text = element_text(size = 12),
    axis.title = element_text(size = 15),
    legend.position = c(1, 0.01),
    legend.justification = c(1, 0),
    legend.key.width = grid::unit(35, "pt"),
    )

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'