How to Choose a Chart
A Statistically Motivated Guide
Andy Grogan-Kaylor
August 23, 2021
library(ggplot2) # beautiful graphs
library(ggthemes) # nice themes for ggplot2
library(ggbeeswarm) # helpful 'beeswarm' geometry
library(ggdist) # 'distribution' geometries
library(cowplot) # arrange graphs
library(pander) # nice tables
How to Choose a Chart
Choosing the right chart to represent your data can be a daunting process. I believe that a starting point for this thinking is some basic statistical thinking about the type of variables that you have. At the broadest level, variables may be conceptualized as categorical variables, or continuous variables.
- categorical variables represent unordered categories like gender, or religious affiliation.
- continuous variables represent a continuous scale like a mental health scale, or a measure of neighborhood quality.
Once we have discerned the type of variable that have, there are two followup questions we may ask before deciding upon a chart strategy:
- Is our graph about one thing at a time?
- How much of x is there?
- What is the distribution of x?
- Is our graph about two things at a time?
- What is the relationship of x and y?
- How are x and y associated?
A Few Notes
A Note About Software
The principles of graphing discussed in this document transcend any particular software package, and could be implemented in many different software packages, such as SPSS, SAS, Stata, or R.
The graphs in these particular examples use ggplot2, a graphing library in R. ggplot2 graph syntax can be formidably complex, with a steep learning curve. More information about ggplot can be found here.
A Note About The Code In This Document
Note that ggplot2 can be MUCH simpler than these examples make it look.
For example,
ggplot(mydata, aes(x = x)) + geom_histogram()
will produce a perfectly serviceable histogram.
Much of the complication of the code in this document is simply the result of formatting tweaks to get the graphs EXACTLY the way I wanted them.
Observe also, that for layout purposes, I am reading each ggplot call into an object, e.g.
p1 <- ggplot(mydata, aes(x = x)) + geom_histogram()
so that I can later use plot_grid to lay out the graphs.
In your own work, you do not need to do this, and it may be simpler to simply say:
ggplot(...) + ...
A Note About Graph Colors
This document uses colors based upon official University of Michigan colors. Using colors that match the design scheme of your organization may be helpful.
# michigan colors
michigan_colors=c("#00274c", # blue
"#ffcb05", # maize
"#a4270b", # tappan red
"#e96300", # ross school orange
"#beb300", # wave field green
"#21c1bc", # taubman teal
"#2878ba", # arboretum blue
"#7207a5") # ann arbor amethyst
# name individual colors
michigan_blue <- "#00274c"
michigan_maize <- "#ffcb05"
tappan_red <- "#a4270b"
ross_school_orange <- "#e96300"
wave_field_green <- "#beb300"
taubman_teal <- "#21c1bc"
arboretum_blue <- "#2878ba"
ann_arbor_amethyst <- "#7207a5"A Simulated Data File of Continuous and Categorical Data
A few randomly selected observations…
| Â | x | y | z | u | v | w | s | q |
|---|---|---|---|---|---|---|---|---|
| 983 | 96.46 | 152 | 106 | Group B | Group B | Group A | Group 2 | 116.5 |
| 557 | 240.7 | 181.2 | 77.56 | Group A | Group A | Group B | Group 2 | 260.7 |
| 43 | 158.5 | 411.7 | 125 | Group A | Group A | Group B | Group 2 | 178.5 |
| 828 | 235.8 | 169.5 | 65.58 | Group B | Group B | Group B | Group 3 | 265.8 |
| 217 | 112.8 | 127.8 | 103 | Group B | Group B | Group A | Group 3 | 142.8 |
| 979 | 153.5 | 193.5 | 96.83 | Group B | Group B | Group B | Group 2 | 173.5 |
| 594 | 82.44 | 32.71 | 75.58 | Group A | Group A | Group A | Group 2 | 102.4 |
| 225 | 110.4 | 131.6 | 117.1 | Group B | Group B | Group A | Group 2 | 130.4 |
| 750 | 81.92 | 42.23 | 140.6 | Group B | Group A | Group A | Group 4 | 121.9 |
| 635 | 95.33 | 103.8 | 100.1 | Group A | Group A | Group A | Group 2 | 115.3 |
One Thing At A Time      Two Things At A Time
Continuous      Continuous By Categorical
my_histogram <- ggplot(mydata, aes(x = x)) +
geom_histogram(fill = arboretum_blue) +
ggtitle("histogram") +
xlab("continuous") + ylab("count") +
theme_minimal()
my_facet_histogram <- ggplot(mydata, aes(x = x)) +
geom_histogram(fill = arboretum_blue) +
facet_wrap(~w, nrow = 2) +
ggtitle("histogram by group") +
xlab("continuous") + ylab("count") +
theme_minimal() +
theme(axis.text = element_text(size = 5)) # small font size for axis
plot_grid(my_histogram, my_facet_histogram, ncol=2)
my_density <- ggplot(mydata, aes(x = y)) +
geom_density(fill = michigan_maize) +
ggtitle("density") +
xlab("continuous") + ylab("density") +
theme_minimal()
my_facet_density <- ggplot(mydata, aes(x = y)) +
geom_density(fill = michigan_maize) +
facet_wrap(~w, nrow = 2) +
ggtitle("density by group") +
xlab("continuous") + ylab("density") +
theme_minimal() +
theme(axis.text = element_text(size = 5)) # small font size for axis
plot_grid(my_density, my_facet_density, ncol = 2)
my_m_barchart <- ggplot(mydata,
aes(x = 1,
y = q,
fill = factor(1))) +
stat_summary(fun = mean, geom = "bar") +
scale_fill_manual(values = c(arboretum_blue)) +
ggtitle("barchart of mean") +
guides(fill=FALSE) +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.ticks.x = element_blank())
my_facet_m_barchart <- ggplot(mydata,
aes(x = factor(s),
y = q,
fill = s)) +
stat_summary(fun = mean, geom = "bar") +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("barchart of mean \nby group") +
guides(fill=FALSE) +
xlab("categorical") + ylab("mean of continuous") +
theme_minimal()
plot_grid(my_m_barchart,
my_facet_m_barchart,
ncol = 2)
my_horiz_m_barchart <- ggplot(mydata,
aes(x = 1,
y = q,
fill = factor(1))) +
stat_summary(fun = mean,
geom = "bar") +
coord_flip() +
scale_fill_manual(values = c(arboretum_blue)) +
ggtitle("horizontal barchart of mean") +
guides(fill=FALSE) +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.y = element_blank()) +
theme(axis.ticks.y = element_blank())
my_facet_horiz_m_barchart <- ggplot(mydata,
aes(x = factor(s),
y = q,
fill = s)) +
stat_summary(fun = mean,
geom = "bar") +
coord_flip() +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("horizontal barchart of mean \nby group") +
guides(fill=FALSE) +
xlab(" ") + ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.y = element_blank()) +
theme(axis.ticks.y = element_blank())
plot_grid(my_horiz_m_barchart,
my_facet_horiz_m_barchart)
my_horiz_m_dotchart <- ggplot(mydata,
aes(x = 1,
y = q,
fill = factor(1))) +
stat_summary(fun = mean,
geom = "point", size = 5) +
coord_flip() +
scale_color_manual(values = c(arboretum_blue)) +
ggtitle("horizontal dotchart of mean") +
guides(fill = FALSE) +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.y = element_blank(),
axis.ticks.y = element_blank())
my_facet_horiz_m_dotchart <- ggplot(mydata,
aes(x = factor(s),
y = q,
color = s)) +
stat_summary(fun = mean,
geom = "point",
size = 5) +
coord_flip() +
scale_color_manual(name = "group",
values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("horizontal dotchart of mean \nby group") +
guides(fill=FALSE) +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.title.y = element_blank(),
axis.ticks = element_blank())
plot_grid(my_horiz_m_dotchart,
my_facet_horiz_m_dotchart)
my_horiz_m_lollipop_chart <- ggplot(mydata,
aes(x = 1,
y = q,
fill = factor(1))) +
stat_summary(fun = mean,
geom = "point",
size = 5) +
geom_segment(aes(x = 1,
xend = 1,
y = 0,
yend = mean(q))) +
coord_flip() +
scale_color_manual(values = c(arboretum_blue)) +
ggtitle("horizontal lollipop chart of mean") +
guides(fill = FALSE) +
xlab(" ") + ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.y = element_blank(),
axis.ticks.y = element_blank())
my_facet_horiz_m_lollipop_chart <- ggplot(mydata,
aes(x = factor(s),
y = q,
color = s)) +
stat_summary(fun = mean,
geom = "point",
size = 5) +
geom_segment(aes(x = factor(s),
xend = factor(s),
y = 0,
yend = mean(q))) +
coord_flip() +
scale_color_manual(name = "group",
values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("horizontal lollipop chart of mean \nby group") +
guides(fill=FALSE) +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.title.y = element_blank(),
axis.ticks = element_blank())
plot_grid(my_horiz_m_lollipop_chart,
my_facet_horiz_m_lollipop_chart)
my_m_linechart <- ggplot(mydata,
aes(x = factor(s),
y = mean(q),
group = 1)) +
stat_summary(fun = mean,
geom = "line",
size = 2,
color = arboretum_blue) +
geom_blank() +
ggtitle("linechart of mean") +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.ticks.x = element_blank())
my_facet_m_linechart <- ggplot(mydata,
aes(x = factor(s),
y = q,
group = 1)) +
stat_summary(fun = mean,
geom = "line",
size = 2,
color = arboretum_blue) +
ggtitle("linechart of mean \nby group") +
xlab(" ") +
ylab("mean of continuous") +
theme_minimal()
plot_grid(my_m_linechart, my_facet_m_linechart)
my_violin <- ggplot(mydata,
aes(x = 1,
y = y)) +
geom_violin(fill = ross_school_orange) +
ggtitle("violin plot") +
xlab(" ") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.ticks.x = element_blank())
my_facet_violin <- ggplot(mydata,
aes(x = 1,
y = y)) +
geom_violin(fill = ross_school_orange) +
facet_wrap(~w,
ncol = 2) +
ggtitle("violin plot \nby group") +
xlab("categorical") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.ticks.x = element_blank())
plot_grid(my_violin, my_facet_violin, ncol = 2)
my_boxplot <- ggplot(mydata,
aes(x = 2,
y = y)) +
geom_boxplot(colour=tappan_red) +
scale_x_discrete(limit = c(0,1,2)) +
ggtitle("boxplot") +
xlab(" ") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.ticks.x = element_blank())
my_conditional_boxplot <- ggplot(mydata,
aes(x = w,
y = y)) +
geom_boxplot(colour=tappan_red, width = .5) +
ggtitle("boxplot \nby group") +
xlab("categorical") +
ylab("continuous") +
theme_minimal()
plot_grid(my_boxplot, my_conditional_boxplot, ncol = 2)
library(ggbeeswarm) # beeswarm geometry
my_beeswarm <- ggplot(mydata,
aes(x = y,
y = 1)) +
geom_beeswarm(colour = ann_arbor_amethyst,
groupOnX = FALSE) +
ggtitle("beeswarm plot") +
xlab("continuous") +
ylab("") +
theme_minimal() +
theme(axis.text=element_text(size = 5),
axis.text.y = element_blank())
my_facet_beeswarm <- ggplot(mydata,
aes(x = y,
y = 1)) +
geom_beeswarm(colour = ann_arbor_amethyst,
groupOnX = FALSE) +
facet_wrap(~w, nrow = 2) +
ggtitle("beeswarm plot \nby group") +
xlab("continuous") +
ylab("") +
theme_minimal() +
theme(axis.text.x = element_text(size = 5),
axis.text.y = element_blank())
plot_grid(my_beeswarm, my_facet_beeswarm)
library(ggdist) # 'distribution' geometries
my_dotplot <- ggplot(mydata,
aes(x = y)) +
stat_dots(colour=wave_field_green) +
ggtitle("dotplot") +
xlab("continuous") +
ylab("density") +
theme_minimal()
my_facet_dotplot <- ggplot(mydata,
aes(x = y)) +
stat_dots(colour=wave_field_green) +
facet_wrap(~w, nrow = 2) +
ggtitle("dotplot \nby group") +
xlab("continuous") +
ylab("density") +
theme_minimal() +
theme(axis.text = element_text(size = 5)) # small font size for axis
plot_grid(my_dotplot, my_facet_dotplot, ncol = 2)
One Thing At A Time      Two Things At A Time
Categorical      Categorical By Categorical
my_barchart <- ggplot(mydata,
aes(s,
fill = s)) +
geom_bar(width = 1.0) +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("bar chart") +
guides(fill=FALSE) +
xlab("categorical") +
ylab("count") +
theme_minimal()
my_facet_barchart <- ggplot(mydata,
aes(s,
fill = s)) +
geom_bar(width = 1.0) +
facet_wrap(~u, ncol = 2) +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("bar chart \nby group") +
guides(fill=FALSE) +
xlab("categorical") + ylab("count") +
theme_minimal() +
theme(axis.text.x = element_text(size = rel(.6)))
plot_grid(my_barchart, my_facet_barchart, ncol = 2)
my_horiz_barchat <- ggplot(mydata, aes(s, fill = s)) +
geom_bar(width = 1.0) +
coord_flip() +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("horizontal bar chart") +
guides(fill=FALSE) +
xlab("categorical") +
ylab("count") +
theme_minimal()
my_horiz_facet_barchart <- ggplot(mydata, aes(s, fill = s)) +
geom_bar(width = 1.0) +
facet_wrap(~u, ncol = 1) +
coord_flip() +
scale_fill_manual(values = c(arboretum_blue,
taubman_teal,
michigan_blue,
michigan_maize)) +
ggtitle("horizontal bar chart \nby group") +
guides(fill=FALSE) +
xlab("categorical") +
ylab("count") +
theme_minimal() +
theme(axis.text.y = element_text(size = rel(.5)))
plot_grid(my_horiz_barchat, my_horiz_facet_barchart, ncol = 2)
my_pie <- ggplot(mydata, aes(x = 1, fill = v)) +
geom_bar(width = 1) +
coord_polar(theta="y") +
scale_fill_manual(values = michigan_colors) +
ggtitle("pie chart") + guides(fill=FALSE) +
xlab(" ") +
ylab("categorical") +
theme_void() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
my_facet_pie <- ggplot(mydata, aes(x = 1, fill = v)) +
geom_bar(width = 1, position = "fill") +
coord_polar(theta="y") +
facet_wrap(~u, ncol = 2) +
scale_fill_manual(values = michigan_colors) +
ggtitle("pie chart \nby group") +
guides(fill=FALSE) +
xlab(" ") +
ylab("categorical") +
theme_void() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
plot_grid(my_pie, my_facet_pie, ncol = 2)
my_doughnut <- ggplot(mydata, aes(x = 2, fill = v)) +
geom_bar() +
coord_polar(theta="y") +
scale_fill_manual(values = michigan_colors) +
ggtitle("doughnut chart") +
guides(fill=FALSE) +
xlab(" ") +
ylab("categorical") +
xlim(.5, 2.5) +
theme_void() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
my_facet_doughnut <- ggplot(mydata, aes(x = 2, fill = v)) +
geom_bar(position = "fill") +
coord_polar(theta="y") +
facet_wrap(~u, ncol = 2) +
scale_fill_manual(values = michigan_colors) +
ggtitle("doughnut chart \nby group") +
guides(fill=FALSE) +
xlab(" ") +
ylab("categorical") +
xlim(.5, 2.5) +
theme_void() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
plot_grid(my_doughnut, my_facet_doughnut, ncol = 2)
Continuous by Continuous
my_scatterplot <- ggplot(mydata, aes(x=x, y=y)) +
geom_point(colour = ann_arbor_amethyst) +
ggtitle("scatterplot") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
my_scatterplot_smoother <- ggplot(mydata, aes(x=x, y=y)) +
geom_point(colour = ann_arbor_amethyst) +
geom_smooth(method = lm, color = michigan_maize, size = 2) +
ggtitle("scatterplot with fit line") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
plot_grid(my_scatterplot, my_scatterplot_smoother)
my_hexagon <- ggplot(mydata, aes(x=x, y=y)) +
geom_hex() +
ggtitle("hexagon plot") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(legend.text = element_text(size=4),
legend.key.size = unit(.25, "cm")) +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank()) +
scale_fill_gradient(low = taubman_teal,
high = arboretum_blue)
my_smoother <- ggplot(mydata, aes(x=x, y=y)) +
geom_point(colour = ann_arbor_amethyst) +
geom_smooth(se=TRUE, color=michigan_maize, size=2) +
ggtitle("scatterplot with smoother") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
plot_grid(my_hexagon, my_smoother)
my_area <- ggplot(mydata, aes(x=x, y=y)) +
geom_area(position = "stack", fill = ross_school_orange) +
ggtitle("area plot") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank())
my_contour <- ggplot(mydata, aes(x=x, y=y)) +
stat_density_2d(aes(fill = ..level..),
geom = "polygon") +
ggtitle("contour plot") +
xlab("continuous") +
ylab("continuous") +
theme_minimal() +
theme(axis.text.x = element_blank()) +
theme(axis.text.y = element_blank()) +
theme(axis.ticks = element_blank()) +
scale_fill_gradient(low = taubman_teal,
high = arboretum_blue)
plot_grid(my_area, my_contour)
Graphics made with the ggplot2 graphing library created by Hadley Wickham.
Available online at https://agrogan1.github.io/
How to Choose a Chart by Andrew Grogan-Kaylor is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. You are welcome to download and use this handout in your own classes, or work, as long as the handout remains properly attributed.
Last updated: August 23 2021 at 09:36