Information design and data visualization
Week 3, Lecture 06
Richard E.W. Berl
Spring 2019
ggplot2
library(ggplot2)Load the egg data as a CSV.
egg = read.csv("./data/aaj1945_DataS1_Egg_shape_by_species_v2.csv", header=T,
stringsAsFactors=F)
colnames(egg)[7:9] = c("AvgLength","NumberOfImages","NumberOfEggs")
egg = egg[-c(1401,1402),]str(egg)## 'data.frame': 1400 obs. of 9 variables:
## $ Order : chr "ACCIPITRIFORMES" "ACCIPITRIFORMES" "ACCIPITRIFORMES" "ACCIPITRIFORMES" ...
## $ Family : chr "Accipitridae" "Accipitridae" "Accipitridae" "Accipitridae" ...
## $ MVZDatabase : chr "Accipiter badius" "Accipiter cooperii" "Accipiter gentilis" "Accipiter nisus" ...
## $ Species : chr "Accipiter badius" "Accipiter cooperii" "Accipiter gentilis" "Accipiter nisus" ...
## $ Asymmetry : num 0.1378 0.0937 0.1114 0.0808 0.0749 ...
## $ Ellipticity : num 0.344 0.272 0.319 0.239 0.254 ...
## $ AvgLength : num 3.86 4.9 5.99 4.04 3.87 ...
## $ NumberOfImages: int 1 27 7 13 15 1 191 1 7 2 ...
## $ NumberOfEggs : int 2 103 18 61 57 1 391 2 17 4 ...Subset the full data frame to Order Galliformes, chickens and other fowl.
eggGalli = egg[egg$Order == "GALLIFORMES",]head(eggGalli)## Order Family MVZDatabase
## 392 GALLIFORMES Cracidae Ortalis vetula
## 393 GALLIFORMES Megapodiidae Macrocephalon maleo
## 394 GALLIFORMES Megapodiidae Megapodius freycinet
## 395 GALLIFORMES Megapodiidae Megapodius pritchardii
## 396 GALLIFORMES Numididae Numida meleagris
## 397 GALLIFORMES Odontophoridae Callipepla californica
## Species Asymmetry Ellipticity AvgLength NumberOfImages
## 392 Ortalis vetula 0.1178 0.4590 6.0847 10
## 393 Macrocephalon maleo 0.0365 0.7237 8.3400 1
## 394 Megapodius freycinet 0.0551 0.6199 8.9621 1
## 395 Megapodius pritchardii 0.0221 0.7072 7.9284 1
## 396 Numida meleagris 0.2260 0.3197 5.2572 1
## 397 Callipepla californica 0.2364 0.2850 3.0381 55
## NumberOfEggs
## 392 31
## 393 1
## 394 2
## 395 3
## 396 10
## 397 700Violin plot
ggplot(data=eggGalli, mapping=aes(x=Family, y=Ellipticity)) +
geom_violin()
table(eggGalli$Family)##
## Cracidae Megapodiidae Numididae Odontophoridae Phasianidae
## 1 3 1 7 36Cracidae and Numididae don’t show distributions because they each only have 1 observation.
We can subset the plot to not show them. Alternatively, we could add a geom_point() layer to show them as single points.
ggplot(data=eggGalli[eggGalli$Family != "Cracidae" &
eggGalli$Family != "Numididae",],
mapping=aes(x=Family, y=Ellipticity)) +
geom_violin()
Our axis doesn’t start at zero, which can be misleading.
range(egg$Ellipticity)## [1] 0.0967 0.7237Looks like the range of possible values is probably 0 to 1. Let’s change our axis, and add the expand argument to get rid of the extra space beyond the axis limits.
ggplot(data=eggGalli[eggGalli$Family != "Cracidae" &
eggGalli$Family != "Numididae",],
mapping=aes(x=Family, y=Ellipticity)) +
geom_violin() +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0))
Compare this to a bar plot of the same data:
ggplot(data=eggGalli[eggGalli$Family != "Cracidae" &
eggGalli$Family != "Numididae",],
mapping=aes(x=Family, y=Ellipticity)) +
geom_violin() +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0))
ggplot(data=eggGalli[eggGalli$Family != "Cracidae" &
eggGalli$Family != "Numididae",],
mapping=aes(x=Family, y=Ellipticity)) +
geom_bar(stat="summary", fun.y="mean") +
geom_errorbar(stat="summary", fun.data="mean_cl_normal",
width=0.5) +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0))
If we want, we can take the best of both and lay the confidence intervals on top of the violin plot:
ggplot(data=eggGalli[eggGalli$Family != "Cracidae" &
eggGalli$Family != "Numididae",],
mapping=aes(x=Family, y=Ellipticity)) +
geom_violin() +
geom_errorbar(stat="summary", fun.data="mean_cl_normal",
width=0.1) +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0))
Facets
Plot all ellipticity values by species.
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
geom_bar(stat="identity") +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
Add a facet to group them by family.
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family)) +
geom_bar(stat="identity") +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
… But all of our values appear under each facet. Not what we want. So we add scales="free" to allow scales to vary across rows and columns.
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free") +
geom_bar(stat="identity") +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
Still not quite right, as some bars are now wider than others and may seem like they’re communicating some other information. So we add space="free" to allow panel sizes to vary.
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
And we’ve now made a new problem in that our facet labels don’t fit anymore. So let’s use the recode() function from dplyr to change those values.
unique(eggGalli$Family)## [1] "Cracidae" "Megapodiidae" "Numididae" "Odontophoridae"
## [5] "Phasianidae"library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unioneggGalli$Family = recode(eggGalli$Family,
Cracidae = "C",
Megapodiidae = "M",
Numididae = "N",
Odontophoridae = "O",
Phasianidae = "P")ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
Themes
theme_gray() is the default
theme_minimal()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
theme_bw()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_bw() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
theme_dark()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_dark() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
theme_void()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_void() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
Others listed here: https://ggplot2.tidyverse.org/reference/ggtheme.html
Some themes are also available in external packages, such as ggthemes.
install.packages("ggthemes")library(ggthemes)## Warning: package 'ggthemes' was built under R version 3.5.3ggthemes::theme_wsj()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_wsj() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
ggthemes::theme_fivethirtyeight()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_fivethirtyeight() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
ggthemes::theme_economist()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_economist() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
ggthemes::theme_tufte()
ggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_tufte() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7))
Or, for fun, xkcd… (like the webcomic)
install.packages("xkcd")
install.packages("extrafont")library(xkcd)## Loading required package: extrafont## Registering fonts with Rdownload.file("http://simonsoftware.se/other/xkcd.ttf",
dest="./resources/xkcd.ttf", mode="wb")
font_import(paths="./resources/", pattern="[X/x]kcd", prompt=FALSE)
loadfonts(device="win") # For Mac: loadfonts()
# Also have to put the font in your system fonts directoryggplot(data=eggGalli, mapping=aes(x=Species, y=Ellipticity)) +
facet_grid(cols=vars(Family), scales="free", space="free") +
geom_bar(stat="identity") +
labs(title="Egg ellipticity by family in Galliformes spp.",
caption="Data from Stoddard et al. (2017)") +
theme_xkcd() +
theme(axis.text.x=element_text(angle=50, hjust=1, size=7),
text=element_text(family="xkcd"))
There are additional functions to make “fuzzy” rectangles and lines, but they’re very particular about the input so you’d have to reformat your data:
?xkcdrect
?xkcdlineColor scales
Arrange all species in the original data frame by their ellipticity values (and remove the grid and tick marks):
ggplot(data=egg, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0)) +
theme_minimal() +
theme(axis.text.x=element_blank(),
panel.grid.major=element_blank(),
panel.grid.minor=element_blank())## Warning: Removed 2 rows containing missing values (geom_bar).
We’re seeing a warning about missing values.
egg[is.na(egg$Species),]## Order Family MVZDatabase Species Asymmetry
## 377 EXTINCT Aepyornithidae Aepyornis sp. <NA> 0.0044
## 378 EXTINCT Columbidae Ectopistes migratorius <NA> 0.0872
## 379 EXTINCT Dinornithidae Dinornis sp. <NA> 0.0050
## 380 EXTINCT Rallidae Porzana palmeri <NA> 0.0523
## Ellipticity AvgLength NumberOfImages NumberOfEggs
## 377 0.4499 23.8700 2 2
## 378 0.3391 3.9290 5 6
## 379 0.3318 14.4235 1 1
## 380 0.4494 2.9990 1 1There are some extinct species without a Species value. To get rid of the warning, we can fix it for now by using the MVZDatabase values, which are good enough:
egg$Species[is.na(egg$Species)] = egg$MVZDatabase[is.na(egg$Species)]
egg[is.na(egg$Species),]## [1] Order Family MVZDatabase Species
## [5] Asymmetry Ellipticity AvgLength NumberOfImages
## [9] NumberOfEggs
## <0 rows> (or 0-length row.names)We also saw a beautiful sigmoidal curve for the ellipticity values when we lined them all up. Why?
ggplot(data=egg, mapping=aes(x=Ellipticity)) +
geom_density()
Because the distribution of values is really close to a “normal” Gaussian curve.
Now let’s subset the data for Order Cuculiformes, the cuckoos.
eggCuckoo = egg[egg$Order == "CUCULIFORMES",]
eggCuckoo$Species## [1] "Centropus bengalensis" "Centropus senegalensis"
## [3] "Centropus superciliosus" "Centropus viridis"
## [5] "Coccyzus americanus" "Coccyzus erythropthalmus"
## [7] "Coccyzus minor" "Crotophaga ani"
## [9] "Crotophaga sulcirostris" "Cuculus pallidus"
## [11] "Geococcyx californianus" "Geococcyx velox"
## [13] "Guira guira" "Phaenicophaeus superciliosus"We also want to take these Species values and split them to make a Genus variable too (since we have Order, Family, and Species, but not Genus).
We do this by using strsplit() and a fancy apply()-family function:
strsplit(eggCuckoo$Species, " ")## [[1]]
## [1] "Centropus" "bengalensis"
##
## [[2]]
## [1] "Centropus" "senegalensis"
##
## [[3]]
## [1] "Centropus" "superciliosus"
##
## [[4]]
## [1] "Centropus" "viridis"
##
## [[5]]
## [1] "Coccyzus" "americanus"
##
## [[6]]
## [1] "Coccyzus" "erythropthalmus"
##
## [[7]]
## [1] "Coccyzus" "minor"
##
## [[8]]
## [1] "Crotophaga" "ani"
##
## [[9]]
## [1] "Crotophaga" "sulcirostris"
##
## [[10]]
## [1] "Cuculus" "pallidus"
##
## [[11]]
## [1] "Geococcyx" "californianus"
##
## [[12]]
## [1] "Geococcyx" "velox"
##
## [[13]]
## [1] "Guira" "guira"
##
## [[14]]
## [1] "Phaenicophaeus" "superciliosus"eggCuckoo$Genus = sapply(strsplit(eggCuckoo$Species, " "), "[", 1)
unique(eggCuckoo$Genus)## [1] "Centropus" "Coccyzus" "Crotophaga" "Cuculus"
## [5] "Geococcyx" "Guira" "Phaenicophaeus"Now let’s plot the cuckoo species, ordered by ellipticity values, with their fill colors according to genus.
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,1,0.2),
limits=c(0,1),
expand=c(0,0)) +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Let’s readjust the Y axis and fix the X axis label.
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Now we want to customize the colors instead of using the default R palette.
length(unique(eggCuckoo$Genus))## [1] 7We will need 7 different color values, and those values need to be for discrete or categorical data, not continuous (Ellipticity, for instance, would need a continuous scale).
Manual
The first option is to assign color values manually, either by name or by hexadecimal value.
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=c("#800000","#9A6324","#808000","#469990",
"#000075","#000000","#e6194B")) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Source: Sasha Trubetskoy
ColorBrewer
Or we can use the brewer scales that are included in ggplot2 and can be viewed online at: http://colorbrewer2.org/
Here, we just specify the type of scale (qualitative) and the palette to be used (here, “Accent”). Other choices are shown in ?scale_fill_brewer.
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_brewer(type="qual", palette="Accent") +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Source: ColorBrewer 2.0
Color-blind friendly
But none of the palettes we’ve used so far are color-blind friendly, and may not be discernable in print or grayscale (after being photocopied), either. There are better palettes out there that are.
dichromat
install.packages("dichromat")library(dichromat)## Warning: package 'dichromat' was built under R version 3.5.2The dichromat() function will take a list of colors (again, named or hexadecimal values) and convert them into a comparable color-blind friendly palette for a specific type of color-blindness (here: protanopia, the kind I have). When in doubt, you can use deuteranopia, the most common form of dichromacy.
dichromat(c("goldenrod","lawngreen","firebrick",
"darksalmon","dodgerblue","darkviolet",
"darkslateblue"),
type="protan")## [1] "#ACAC2A" "#F1F10A" "#4B4B2C" "#A1A17B" "#8888FE" "#3939D3" "#40408B"Compare the original colors (on the left) with the ones substituted by dichromat() (on the right):
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=c("goldenrod","lawngreen","firebrick",
"darksalmon","dodgerblue","darkviolet",
"darkslateblue")) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=dichromat(c("goldenrod","lawngreen","firebrick",
"darksalmon","dodgerblue","darkviolet",
"darkslateblue"),
"protan")) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
dichromat also comes with some predefined color schemes in the colorschemes object. Here, we use colorschemes$Categorical.12 because we’re using categorical data, and subset the first 7 elements for our 7 levels:
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=dichromat::colorschemes$Categorical.12[1:7]) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
I find this one doesn’t work great for me personally.
viridis
install.packages("viridis")library(viridis)## Warning: package 'viridis' was built under R version 3.5.3## Loading required package: viridisLiteThe viridis package has a number of color-blind friendly palettes, listed under their own functions that end with _d (for discrete) or _c (for continuous). The default is "viridis":
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_viridis_d() +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Still not great for me, with all the dark blues/purples.
But there are others, such as "magma":
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_viridis_d(option="magma") +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
Again, not my preference, but it’s there and could work.
Paul Tol
Finally, my absolute favorite:
Tol, P. (2018). Colour schemes. SRON Netherlands Institute for Space Research. Available: https://personal.sron.nl/~pault/data/colourschemes.pdf
The older version of his qualitative color scheme (which I actually like better than the new one) is available in the ggthemes package, which we already have loaded. Weirdly, you have to specify the number of levels in parentheses after the function, rather than as an argument in the function.
ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=ggthemes::ptol_pal()(7)) +
labs(x="Species") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))
A variety of other color schemes are available in the document linked above which, in addition to being color-blind friendly, are also distinct in grayscale.
Saving plots
ggplot()
First, assign your plot to an object:
plotCuckoo = ggplot(data=eggCuckoo, mapping=aes(x=reorder(Species, Ellipticity),
y=Ellipticity,
fill=Genus)) +
geom_bar(stat="identity") +
scale_y_continuous(breaks=seq(0,0.5,0.1),
limits=c(0,0.5),
expand=c(0,0)) +
scale_fill_manual(values=ggthemes::ptol_pal()(7)) +
labs(title="Egg ellipticity in Cuculiformes spp.", x="Species",
caption="Data from Stoddard et al. (2017)") +
theme_minimal() +
theme(axis.text.x=element_text(angle=50, hjust=1))Then give it as an argument to the ggsave() function:
JPG
ggsave("./figs/plotCuckoo.jpg", plotCuckoo, width=8, height=6, units="in",
dpi=300)ggsave("./figs/plotCuckoo.pdf", plotCuckoo, width=8, height=6, units="in")base
These functions let you save any kind of graphic, including ggplots. You have to use the graphic function first to open a “graphic device” (where the plot is sent), then run the code to produce the plot, then close the connection to the device with dev.off().
JPG
jpeg("./figs/plotCuckoo2.jpg", width=8, height=6, units="in",
res=300)
plotCuckoo
dev.off()pdf("./figs/plotCuckoo2.pdf", width=8, height=6)
plotCuckoo
dev.off()