03b. Code for visual encodings notes

These are the charts from the notes on visual encodings, but focused on the code. Otherwise, as always, the code for this whole site is on GitHub.

library(ggplot2)
library(dplyr)

# pull a qualitative palette from carto colors
qual_pal <- rcartocolor::carto_pal(name = "Bold")
# colorspace::swatchplot(qual_pal)
gender_pal <- qual_pal[c(3, 7)]

# pull a diverging palette but make center value gray
div_pal <- rcartocolor::carto_pal(n = 3, name = "Earth")
div_pal[2] <- "gray40"

sex_x_edu <- justviz::wages |>
    filter(dimension %in% c("by_edu", "by_sex_x_edu"), status == "full_time") |>
    select(sex, edu, earn_q20, earn_q50, earn_q80)

sex_x_edu

sex	edu	earn_q20	earn_q50	earn_q80
total	no_diploma	27868	43234	70000
total	high_school_diploma	33227	52520	85018
total	some_college	40000	64311	104451
total	bachelors	54655	91091	145771
total	graduate_degree	72873	116000	181895
men	no_diploma	30881	48582	75145
men	high_school_diploma	36436	58000	92644
men	some_college	42874	72057	115382
men	bachelors	60727	102938	162199
men	graduate_degree	81100	134000	202749
women	no_diploma	21617	35264	57928
women	high_school_diploma	30700	46322	72873
women	some_college	36436	56616	92685
women	bachelors	50000	80160	126000
women	graduate_degree	68015	102938	157891

Palettes are just vectors of colors, usually hex codes. For example, gender_pal is just #3969AC, #E68310. There are functions to view them: you already have colorspace installed, so call colorspace::swatchplot(gender_pal) to view the colors.

colorspace::swatchplot(qual_pal)

Scales only, no geometries: inside ggplot’s aes (aesthestics) call, assign education to the x scale and median earnings to the y scale. Let’s call this chart 1.

sex_x_edu |>
    filter(sex == "total") |>
    ggplot(aes(x = edu, y = earn_q50))

Same as chart 1, but add a point geometry (see ?geom_point)

sex_x_edu |>
    filter(sex == "total") |>
    ggplot(aes(x = edu, y = earn_q50)) +
    geom_point(size = 6)

Same as chart 1, but instead add a column geometry (see ?geom_col). Call this chart 2.

sex_x_edu |>
    filter(sex == "total") |>
    ggplot(aes(x = edu, y = earn_q50)) +
    geom_col(width = 0.8)

Modify chart 2 to also map sex onto the fill aesthetic. Note that there are independent aesthetics for color and fill, and most larger shapes like columns will default to a fill but not a color (color would provide an outline), while most smaller shapes like points and lines have a color only. (If you look at the standard codes for point shapes, some have both a fill and an outline color; this is very useful in some cases but we won’t get into it now.) Tl;dr if you try to set what you think of as color and don’t see it change, it might actually be the fill.

Add a manual fill scale so we can use the palette we pulled out earlier.

sex_x_edu |>
    filter(sex != "total") |>
    ggplot(aes(x = edu, y = earn_q50, fill = sex)) +
    geom_col(width = 0.8) +
    scale_fill_manual(values = gender_pal)

Bars shouldn’t be stacked, so set the position to dodge them, i.e. put them next to each other. Each bar is a combination of x/education and fill/sex. position_dodge2 puts a nice lil gap between the bars at each x value, whereas position_dodge will have them smooshed together.

sex_x_edu |>
    filter(sex != "total") |>
    ggplot(aes(x = edu, y = earn_q50, fill = sex)) +
    geom_col(width = 0.8, position = position_dodge2()) +
    scale_fill_manual(values = gender_pal)

Instead of medians by sex, now show values at each of 3 percentiles. 20th, 50th, and 80th percentiles each have their own column in the original dataset, but if we want to assign percentile to a fill, we need a single variable that has those percentile breaks and another single variable that has values. This is called tidy data, coined by the main dev of ggplot. The tidyr package helps get data into the correct shape for the tidy data/grammar of graphics paradigm.

Just the data tidying step:

ptiles_tidy <- sex_x_edu |>
    filter(sex == "total") |>
    tidyr::pivot_longer(
        -sex:-edu,
        names_to = "percentile",
        values_to = "earnings",
        names_ptypes = list(percentile = factor())
    )
ptiles_tidy

sex	edu	percentile	earnings
total	no_diploma	earn_q20	27868
total	no_diploma	earn_q50	43234
total	no_diploma	earn_q80	70000
total	high_school_diploma	earn_q20	33227
total	high_school_diploma	earn_q50	52520
total	high_school_diploma	earn_q80	85018
total	some_college	earn_q20	40000
total	some_college	earn_q50	64311
total	some_college	earn_q80	104451
total	bachelors	earn_q20	54655
total	bachelors	earn_q50	91091
total	bachelors	earn_q80	145771
total	graduate_degree	earn_q20	72873
total	graduate_degree	earn_q50	116000
total	graduate_degree	earn_q80	181895

Now plot that data with earnings on y and percentile on fill. Call it chart 3

ptiles_tidy |>
    ggplot(aes(x = edu, y = earnings, fill = percentile)) +
    geom_col(width = 0.8, position = position_dodge2()) +
    scale_fill_manual(values = div_pal)

Chart 3 but with points instead of bars. Note that this requires switching from fill to color as our aesthetic.

ptiles_tidy |>
    ggplot(aes(x = edu, y = earnings, color = percentile)) +
    geom_point(size = 6) +
    scale_color_manual(values = div_pal)

Add a path geometry to join the points within each education level and across percentiles. Put it before the point layer so points are on top.

ptiles_tidy |>
    ggplot(aes(x = edu, y = earnings, color = percentile)) +
    geom_path(color = "gray80", linewidth = 3) +
    geom_point(size = 6) +
    scale_color_manual(values = div_pal)

Add a variable for percentile type, whether it’s 50th percentile (median) or one of the others (we’ll call them endpoints, I don’t have a better name). This gives us a column to assign more aesthetics to so medians stand out from 20th & 80th percentiles and have more visual importance.

Now assign point size to percentile type (note that you can do an aes call within any geom; since it only affects points I’ve put it there). Add a manual size scale.

ptiles_endpts <- ptiles_tidy |>
    mutate(
        ptile_type = ifelse(percentile == "earn_q50", "median", "endpoint")
    )

ptiles_endpts

sex	edu	percentile	earnings	ptile_type
total	no_diploma	earn_q20	27868	endpoint
total	no_diploma	earn_q50	43234	median
total	no_diploma	earn_q80	70000	endpoint
total	high_school_diploma	earn_q20	33227	endpoint
total	high_school_diploma	earn_q50	52520	median
total	high_school_diploma	earn_q80	85018	endpoint
total	some_college	earn_q20	40000	endpoint
total	some_college	earn_q50	64311	median
total	some_college	earn_q80	104451	endpoint
total	bachelors	earn_q20	54655	endpoint
total	bachelors	earn_q50	91091	median
total	bachelors	earn_q80	145771	endpoint
total	graduate_degree	earn_q20	72873	endpoint
total	graduate_degree	earn_q50	116000	median
total	graduate_degree	earn_q80	181895	endpoint

ptiles_endpts |>
    ggplot(aes(
        x = edu,
        y = earnings,
        color = percentile
    )) +
    geom_path(color = "gray80", linewidth = 3) +
    geom_point(aes(size = ptile_type)) +
    scale_color_manual(values = div_pal) +
    scale_size_manual(values = c(median = 6, endpoint = 3.5))

Add a third aesthetic to the points: percentile types should get different shapes as well. This requires a group aesthetic to make clear how the paths should be grouped. Look up the shape codes to find IDs of shapes you want. Some of these shapes are filled, so set both fill and color as the same scale (try taking out the aesthetics argument of the color scale). Add a manual shape scale.

ptiles_endpts |>
    ggplot(aes(
        x = edu,
        y = earnings,
        color = percentile,
        fill = percentile,
        group = edu
    )) +
    geom_path(color = "gray80", linewidth = 3) +
    geom_point(aes(size = ptile_type, shape = percentile)) +
    scale_color_manual(values = div_pal, aesthetics = c("color", "fill")) +
    scale_size_manual(values = c(median = 6.5, endpoint = 3.5)) +
    scale_shape_manual(values = c(earn_q20 = 25, earn_q80 = 24, earn_q50 = 19))

Last, clean it up. Walk through this code line by line to see the effects of the changes.

# sentence case + line wrapping
wrap_lbls <- function(x) {
    wrapper <- scales::label_wrap(width = 12)
    x <- snakecase::to_sentence_case(x)
    wrapper(x)
}

ptiles_endpts |>
    mutate(
        percentile = forcats::fct_recode(
            percentile,
            "20th pct" = "earn_q20",
            Median = "earn_q50",
            "80th pct" = "earn_q80"
        )
    ) |>
    ggplot(aes(
        x = edu,
        y = earnings,
        color = percentile,
        fill = percentile,
        group = edu
    )) +
    geom_path(color = "gray80", linewidth = 3) +
    geom_point(aes(size = percentile, shape = percentile)) +
    scale_color_manual(
        values = div_pal,
        aesthetics = c("color", "fill"),
        guide = guide_legend(reverse = TRUE)
    ) +
    scale_size_manual(
        values = c("20th pct" = 3.5, "80th pct" = 3.5, Median = 6.5),
        guide = guide_legend(reverse = TRUE)
    ) +
    scale_shape_manual(
        values = c("20th pct" = 25, "80th pct" = 24, Median = 19),
        guide = guide_legend(reverse = TRUE)
    ) +
    scale_x_discrete(labels = wrap_lbls) +
    scale_y_continuous(
        labels = scales::label_dollar(scale = 1 / 1000, suffix = "K"),
        breaks = seq(0, 2e5, by = 2.5e4),
        limits = c(0, NA),
        expand = expansion(mult = c(0, 0.05))
    ) +
    labs(
        title = "Lower-paid adults with graduate degrees earn about as much as higher-paid adults with no diploma",
        subtitle = "20th percentile, median, and 80th percentile individual earnings,\nMaryland adults ages 25+ by educational attainment, 2024",
        caption = "Source: US Census Bureau American Community Survey 2024 5-year estimates,\npublic use microdata sample via IPUMS",
        x = NULL,
        y = "Earnings",
        color = "Percentile",
        fill = "Percentile",
        shape = "Percentile",
        size = "Percentile"
    ) +
    theme_minimal(base_size = 12) +
    theme(
        plot.title.position = "plot",
        plot.caption.position = "plot",
        axis.ticks = element_blank(),
        panel.grid = element_line(color = "gray85"),
        panel.grid.major.x = element_blank(),
        axis.title.y.left = element_text(
            hjust = 1,
            margin = margin(8, 10, 0, 0, unit = "pt"),
            size = rel(0.8)
        ),
        legend.title = element_text(size = rel(0.8)),
        axis.line.x = element_line(color = "gray50"),
        plot.title = ggtext::element_textbox_simple(
            face = "bold",
            lineheight = 0.9,
            margin = margin_part(b = 7)
        ),
        plot.subtitle = element_text(margin = margin_part(b = 10)),
        plot.caption = element_text(
            hjust = 0,
            size = rel(0.7),
            margin = margin_part(t = 15)
        )
    )