How to add 3rd variable in bar chart for ggplot2? - r

Question

I'm trying to make a bar chart with a 3rd variable (which in this case is "frequency") where the 3rd variable changes the width of the bars (higher frequency = larger width). Obviously I have to figure out the sizing, but that is just aesthetics and I can figure that out later. When I use this code I keep getting the error "position_dodge requires non-overlapping x intervals" and the plot then stacks the bars instead of grouping them. Also (maybe this could help) wondering if there is a way to increase the distance between labels on the x-axis (meaning increase the distance between "Iso", "Transition", "P&R Handler", etc.)
All help appreciated.
library(tidyverse)
library(ggrepel)
percentile_playtype = c(70.10, 41.20, 83.90, 0, 0, 97.30, 40, 0, 49.30, 20.10, 88.90, 91.80,
94.60, 0, 83.60, 86.90, 42, 41.10, 46.90, 0, 81.50, 84.00)
frequency = c(8.5,16.5,53.3,0,0,6,7.2,0,2.1,0.6,5.4,1.9,12.4,0,28,8.1,16,1.9,13.6,0,10.6,6.1)
v1 = sqrt(sqrt(sqrt(frequency)))/10
lowsize <- element_text(size=8)
playtype = c("Iso","Transition","P&R Handler","P&R Roll","Post Up","Spot Up",
"Handoff","Cut","Off Screen","Putbacks","Misc")
Player = rep(c("Trae Young","John Collins"), each=11)
PlayData <- data.frame(percentile_playtype,frequency,playtype,Player)
a1 <- ggplot(PlayData, aes(fill=Player, y=percentile_playtype, x=playtype)) +
geom_bar(position="dodge", stat="identity",width=v1)
a1

Answer 1

If you really want frequency to be mapped to bar width, you need to do it the hard way, and calculate those widths, plotting geom_rect rather than geom_bar. From a visual impact point of view it is better to scale the area of the bars rather than their absolute width:
PlayData$playtype_n <- as.numeric(as.factor(PlayData$playtype))
PlayData$frequency_n <- PlayData$frequency/max(PlayData$frequency) * 0.5 /
(PlayData$percentile_playtype / 100) *
(2 * as.numeric(as.factor(PlayData$Player)) - 3) +
as.numeric(as.factor(PlayData$playtype))
ggplot(PlayData, aes(fill = Player)) +
geom_rect(aes(xmin = playtype_n, xmax = frequency_n, ymin = 0,
ymax = percentile_playtype)) +
scale_x_continuous(breaks = sort(unique(PlayData$playtype_n)),
labels = levels(as.factor(PlayData$playtype))) +
scale_fill_manual(values = c("deepskyblue4", "orange")) +
labs(x = "Play type (area scaled to frequency)", y = "Percentile playtype") +
theme_bw()
Personally, I don't think this looks great, and I'm not convinced it's worth the trouble. Another, easier to understand approach might be to use facets
ggplot(PlayData, aes(fill=frequency, y=percentile_playtype, x=playtype)) +
geom_col(position = "dodge", width=0.75) +
geom_text(aes(label = frequency), vjust = 1.5, color = "white") +
facet_wrap(Player~., ncol = 1) +
scale_fill_viridis_c() +
theme_classic() +
theme(panel.grid.major.y = element_line(color = "gray90"),
strip.background = element_blank(),
strip.text.x = element_text(size = 16),
axis.line.x.bottom = element_line())
Or perhaps a labelled scatter plot using ggrepel:
ggplot(PlayData, aes(percentile_playtype, frequency, color = Player)) +
geom_point() +
geom_text_repel(aes(label = playtype), size = 5) +
scale_color_manual(values = c("deepskyblue4", "orange")) +
theme_bw()

Answer 2

Are you trying to mimic something like a mosaic plot?
percentile_playtype = c(70.10, 41.20, 83.90, 0, 0, 97.30, 40, 0, 49.30, 20.10, 88.90, 91.80,
94.60, 0, 83.60, 86.90, 42, 41.10, 46.90, 0, 81.50, 84.00)
frequency = c(8.5,16.5,53.3,0,0,6,7.2,0,2.1,0.6,5.4,1.9,12.4,0,28,8.1,16,1.9,13.6,0,10.6,6.1)
v1 = sqrt(sqrt(sqrt(frequency)))/10
playtype = c("Iso","Transition","P&R Handler","P&R Roll","Post Up","Spot Up",
"Handoff","Cut","Off Screen","Putbacks","Misc")
Player = rep(c("Trae Young","John Collins"), each=11)
PlayData <- data.frame(percentile_playtype,frequency,playtype,Player)
CGPfunctions::PlotXTabs2(PlayData,
x = playtype,
y = Player,
counts = percentile_playtype,
plottype = "mosaic",
x.axis.orientation = "slant",
sample.size.label = FALSE,
label.text.size = 2)