I want to create a pie chart with anexpanded bar chart of one section. I have looked all over and can't find any examples compleated in R before. I was able to create a version in Microsoft XL to demonstrate what i'm after, where the values and titles have been redacted.
Anyone know how this could be done?
Thanks
Hi, and welcome!
A reproducible example, called a reprex is likely to attract more answers, because it doesn't require retracing the steps you may have already taken, like creating a pie chart and a stack bar chart separately. Do you have anything you can share?
This seemed like an interesting problem. I'm sure there are functions and packages out there that will do a lot of this work, but I reinvented a few wheels just because I wanted explore how to put things together. The result is some working code. It's ugly, a little chaotic, but functional. Hopefully it helps you put together a solution that fits your needs.
First, you will need to run all of this code in your session
#--------------------------------------------------------------------
# Source in all of these functions together.
#
# CONTENTS ----------------------------------------------------------
#
# dual_chart - Master function. It will combine two charts and either
# print the plot object, or return the two plots in a list
# where they can be further customized.
# dual_chart_combine - Does the work of putting the two plots side by side.
# dual_chart_pie - Construct the pie chart.
# dual_chart_bar - Construct the bar chart.
# dual_chart --------------------------------------------------------
# Master function. It will combine two charts and either
# print the plot object, or return the two plots in a list
# where they can be further customized.
dual_chart <- function(data, primary_variable, level_for_detail,
secondary_variable,
ncol = 2, nrow = 1, ...,
plot = TRUE){
plots <- list(pie = dual_chart_pie(DF,
primary_variable,
level_for_detail),
bar = dual_chart_bar(DF,
primary_variable,
level_for_detail,
secondary_variable))
if (plot){
dual_chart_combine(plots$pie,
plots$bar,
ncol = ncol,
nrow = nrow,
...)
} else {
plots
}
}
# dual_chart_combine ------------------------------------------------
# Does the work of putting the two plots side by side.
dual_chart_combine <- function(pie_chart, bar_chart,
ncol = 2, nrow = 1, ...){
gridExtra::grid.arrange(
grobs = list(pie_chart, bar_chart),
ncol = ncol,
nrow = nrow
)
}
# dual_chart_pie ----------------------------------------------------
# Construct the pie chart
dual_chart_pie <- function(data, primary_variable, level_for_detail){
require(dplyr)
require(ggplot2)
# Frequency and Relative Frequencies of the data
primary_summary <-
data %>%
group_by_at(primary_variable) %>%
summarise_at(primary_variable,
.funs = list(n = function(x) sum(!is.na(x)))) %>%
mutate_at("n",
.funs = list(p = function(x) x / sum(x)))
# Determine the radian shift.
# This allows us to place the level_for_detail on the right side of the
# pie chart, with the slice of the pie centered over the pi/2 position.
# label_position determines the y-coordinate for the proportion label.
cum_freq <- c(0, cumsum(primary_summary$p))
radian_position <- cum_freq * 6.283
radian_shift <- numeric(length(cum_freq) - 1)
names(radian_shift) <- primary_summary[[primary_variable]]
label_position <- numeric(length(cum_freq) - 1)
for (i in seq_along(radian_shift)){
radian_shift[i] <- mean(radian_position[c(i, i+1)])
label_position[i] <- mean(cum_freq[c(i, i + 1)])
}
# Make the pie chart
ggplot(data = primary_summary,
mapping = aes_string(y = "p",
fill = primary_variable)) +
geom_bar(width = 1,
stat = "identity",
mapping = aes(x = "")) +
geom_text(mapping = aes_string(y = "1 - label_position",
label = "p"),
x = 1) +
coord_polar("y",
start = pi / 2 + radian_shift[level_for_detail])
}
# dual_chart_bar ----------------------------------------------------
# Construct the bar chart
dual_chart_bar <- function(data, primary_variable, level_for_detail,
secondary_variable){
secondary_summary <-
data %>%
group_by_at(c(primary_variable, sub_variable)) %>%
summarise_at(sub_variable,
.funs = list(n = function(x) sum(!is.na(x)))) %>%
ungroup() %>%
mutate_at("n",
.funs = list(p = function(x) x / sum(x))) %>%
filter_at(primary_variable, any_vars(. %in% level_for_detail))
# label_position determines the y-coordinate for the proportion label.
cum_freq <- c(0, cumsum(secondary_summary$p))
label_position <- numeric(length(cum_freq) - 1)
for (i in seq_along(label_position)){
label_position[i] <- mean(cum_freq[c(i, i + 1)])
}
# Make the bar plot
ggplot(data = secondary_summary,
mapping = aes_string(y = "p",
fill = sub_variable)) +
geom_bar(width = 1,
stat = "identity",
mapping = aes(x = "")) +
geom_text(mapping = aes_string(y = "max(cum_freq) - label_position",
label = "p"),
x = 1)
}
Once that's all in there, you can get the crude plot with
dual_chart(DF, "pie_var", "C", "bar_var")
Here's an approach relying more on drawing the elements in a single plot space. This implementation isn't particularly general but might be okay if the output is a one-off.
First, some fake data:
my_data <- data.frame(
stringsAsFactors = F,
Section = c("a", "b", "b", "b", "c"),
Detail = letters[22:26],
Amount = c(0.51, 0.04, 0.05, 0.07, 0.33)
)
Now, loading some libraries and noting the coordinates of the wedge edges.
library(ggplot2); library(ggforce); library(dplyr)
my_data_aug <- my_data %>%
mutate(arc_start = cumsum(lag(Amount, default = 0)) * 2*pi - 2,
arc_end = cumsum(Amount) * 2*pi - 2,
x_pos = 0 + cos(arc_start - pi/2),
y_pos = 1 - sin(arc_start - pi/2))
The detail for the bar chart section:
my_data_detail <- my_data_aug %>%
filter(Section == "b") %>%
purrr::map_df(rev) %>% # This reverses the order
mutate(Amount_scaled = Amount / sum(Amount) * 2)
The detail for the segment lines:
my_data_lines <- my_data_aug %>%
filter(Section == "b" | lag(Section == "b")) %>%
slice(1, n())
Now, all together:
ggplot(my_data_aug) +
geom_arc_bar(aes(x0 = 0, y0 = 1,
r0 = 0, r = 1,
fill = Section,
start = arc_start,
end = arc_end), color = NA) +
geom_tile(data = my_data_detail,
aes(x = 2,
y = cumsum(Amount_scaled) - Amount_scaled/2,
height = Amount_scaled, fill = Detail)) +
annotate("segment",
x = my_data_lines[1:2, "x_pos"],
y = my_data_lines[1:2, "y_pos"],
xend = 1.5,
yend = c(2,0)) +
coord_equal() +
theme_void()
Extra points for omitting the talk!
This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.