Leftover tidbits
Week 13
Introduction
Today we are going to go over a bunch of stuff I thought was interesting but didnβt fit specifically into any of the other lessons. This includes some cool ggplot extension packages we havenβt gone over yet, and heatmaps that utilize base R plotting.
Load libraries
Loading the libraries that are for each section. Individual libraries are before each section so you can see which go with what plot types.
library(tidyverse) # for everything
ββ Attaching core tidyverse packages ββββββββββββββββββββββββ tidyverse 2.0.0 ββ
β dplyr 1.1.3 β readr 2.1.4
β forcats 1.0.0 β stringr 1.5.1
β ggplot2 3.4.4 β tibble 3.2.1
β lubridate 1.9.2 β tidyr 1.3.0
β purrr 1.0.2
ββ Conflicts ββββββββββββββββββββββββββββββββββββββββββ tidyverse_conflicts() ββ
β dplyr::filter() masks stats::filter()
β dplyr::lag() masks stats::lag()
βΉ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
Really start using an Rproject π½οΈ
I have noticed that many of you are still not using RProjects. I would really recommend that for easy file management that you do. Here is an a chapter in R for Data Science on how to set one up. If you want to start using Git in the future, you will need to set up a project.
gghighlight π¦
The package gghighlight
allows you to highlight certain geoms in ggplot. Doing this helps your reader focus on the thing you want them to, and helps prevent plot spaghetti. To practice with gghighlight
we are going to use some data from the R package gapminder
Install
installl.packages("gghighlight")
install.packages("gapminder")
Load libraries
First letβs load our libraries.
library(gghighlight) # for highlighting
library(gapminder) # where data is
Wrangle
We can create a dataframe that includes only the data for the countries in the continent Americas.
<- gapminder %>%
gapminder_americas filter(continent == "Americas")
Plot
If we look at all the countries at once, we get plot spaghetti π.
%>%
gapminder_americas ggplot(aes(x = year, y = lifeExp, group = country, color = country)) +
geom_line() +
theme_minimal() +
labs(x = "Year",
y = "Life Expectancy (years)",
title = "Life Expectancy in Countries in the Americas",
subtitle = "From 1952 to 2007",
caption = "Data from gapminder.org")
Create a lineplot showing the life expectacy over 1952 to 2007 for all countries, highlighting the United States.
# highlight just the US
%>%
gapminder_americas ggplot(aes(x = year, y = lifeExp, group = country, color = country)) +
geom_line() +
gghighlight(country == "United States") +
theme_minimal() +
labs(x = "Year",
y = "Life Expectancy (years)",
title = "Life Expectancy in Countries in the Americas",
subtitle = "From 1952 to 2007",
caption = "Data from gapminder.org")
Facet our plot, and highlight the country for each facet.
# facet and highlight each country
%>%
gapminder_americas ggplot(aes(x = year, y = lifeExp)) +
geom_line(aes(color = country)) +
gghighlight() +
theme_minimal() +
theme(legend.position = "none",
strip.text.x = element_text(size = 8),
axis.text.x = element_text(angle = 90)) +
facet_wrap(vars(country)) +
labs(x = "Year",
y = "Life Expectancy (years)",
title = "Life Expectancy in Countries in the Americas",
subtitle = "From 1952 to 2007",
caption = "Data from gapminder.org")
patchwork, a little more πππ
We have talked a bit about patchwork
in the lecture on PCA but its such a useful package I wanted to go over it a bit more. The goal of patchwork
is to make it very simple to combine plots together.
Load libraries
library(patchwork)
library(palmerpenguins) # for making some plots to assemble
Make some plots
<- penguins %>%
plot1 ggplot(aes(x = species, y = body_mass_g, color = species)) +
geom_boxplot()
<- penguins %>%
plot2 ggplot(aes(x = bill_length_mm, y = bill_depth_mm, color = species)) +
geom_point()
<- penguins %>%
plot3 drop_na() %>%
ggplot(aes(x = island, y = flipper_length_mm, color = species)) +
geom_boxplot() +
facet_wrap(vars(sex))
Combine plots
The simplest ways to combine plots is with the plus sign operator +
. The forward slash /
stacks plots. The pipe |
puts plots next to each other. You can learn more about using patchwork here.
+ plot2) / plot3 (plot1
You can also add annotation and style to your plots. Learn more here.
+ plot2) / plot3 + plot_annotation(tag_levels = c("1"),
(plot1 title = "Here is some information about penguins")
gganimate π
https://gganimate.com/reference/transition_states.html
Install
install.packages("gganimate") # gganimate
install.packages("gapminder") # gapminder data for example
install.packages("magick") # for gif rendering
Load libraries
library(gganimate)
library(ggrepel) # for text/label repelling
library(magick) # for gif rendering
Linking to ImageMagick 6.9.12.3
Enabled features: cairo, fontconfig, freetype, heic, lcms, pango, raw, rsvg, webp
Disabled features: fftw, ghostscript, x11
Plot
First letβs make a base plot. Note that this measure of population isnβt actually correct as its summing all of the populations across all the years.
<- gapminder %>%
(base_plot filter(continent == "Africa") %>%
ggplot(aes(x = pop, y = reorder(country, pop))) +
geom_col() +
scale_x_continuous(labels = scales::unit_format(unit = "M", scale = 1e-6)) +
theme_classic() +
labs(title = "Population from 1952 to 2007 in Africa",
x = "Population",
y = "Country"))
<- base_plot +
(plot_to_animate labs(subtitle = "Year: {frame_time}") + # label subtitle with year
transition_time(year) + # gif over year
ease_aes()) # makes the transitions smoother
# set parameters for your animation
<- animate(plot = plot_to_animate,
animated_plot duration = 10, # number of seconds for whole animation
fps = 10, # framerate, frames/sec
start_pause = 20, # show first time for 20 frames
end_pause = 20, # show end for 20 frames
width = 700, # width in pixels
height = 700, # height in pixels
renderer = magick_renderer()) # program for rendering
Print your animation.
animated_plot
Save
Save your animation.
# save it
anim_save(filename = "gapminder_gif.gif",
animation = last_animation())
ggradar π‘
The package ggradar
allows you to create radar plots, which allow the plotting of multidimensional data on a two dimension chart. Typically with these plots, the goal is to compare the variables on the plot across different groups. We are going to try this out with the coffee tasting data from the distributions recitation.
Install ggradar
if you donβt already have it. This package is not available on CRAN for the newest version of R, so we can use devtools
and install_github()
to install it. You could also try using install.packages()
and see if that works for you.
::install_github("ricardo-bion/ggradar",
devtoolsdependencies = TRUE)
library(ggradar)
library(scales) # for scaling data
# load coffee data from distributions recitation
<- tidytuesdayR::tt_load('2020-07-07') tuesdata
Downloading file 1 of 1: `coffee_ratings.csv`
# extract out df on coffee_ratings
<- tuesdata$coffee_ratings
coffee
# what are the column names again?
colnames(coffee)
[1] "total_cup_points" "species" "owner"
[4] "country_of_origin" "farm_name" "lot_number"
[7] "mill" "ico_number" "company"
[10] "altitude" "region" "producer"
[13] "number_of_bags" "bag_weight" "in_country_partner"
[16] "harvest_year" "grading_date" "owner_1"
[19] "variety" "processing_method" "aroma"
[22] "flavor" "aftertaste" "acidity"
[25] "body" "balance" "uniformity"
[28] "clean_cup" "sweetness" "cupper_points"
[31] "moisture" "category_one_defects" "quakers"
[34] "color" "category_two_defects" "expiration"
[37] "certification_body" "certification_address" "certification_contact"
[40] "unit_of_measurement" "altitude_low_meters" "altitude_high_meters"
[43] "altitude_mean_meters"
We are going to wrangle the data to facilitate plotting. We are using rescale
as we need the data for each attribute to be between 0 and 1.
# tidy data to summarize easily
<- coffee %>%
(coffee_summary_long select(species, aroma:cupper_points) %>% # first column is the groups
pivot_longer(cols = aroma:cupper_points, # our favorite - tidy data to faciliate summarizing
names_to = "attribute",
values_to = "score") %>%
group_by(species, attribute) %>% # perform operations by species and attribute pairs
mutate(across(where(is.numeric), rescale)) %>% # rescale data that is numeric
summarize(mean_score = mean(score)))
`summarise()` has grouped output by 'species'. You can override using the
`.groups` argument.
# A tibble: 20 Γ 3
# Groups: species [2]
species attribute mean_score
<chr> <chr> <dbl>
1 Arabica acidity 0.861
2 Arabica aftertaste 0.853
3 Arabica aroma 0.864
4 Arabica balance 0.859
5 Arabica body 0.876
6 Arabica clean_cup 0.983
7 Arabica cupper_points 0.750
8 Arabica flavor 0.851
9 Arabica sweetness 0.990
10 Arabica uniformity 0.983
11 Robusta acidity 0.707
12 Robusta aftertaste 0.746
13 Robusta aroma 0.603
14 Robusta balance 0.833
15 Robusta body 0.766
16 Robusta clean_cup 0.893
17 Robusta cupper_points 0.507
18 Robusta flavor 0.681
19 Robusta sweetness 0.575
20 Robusta uniformity 0.857
ggradar
takes wide data though, so we are going to pivot back to wide data.
# go back to wide
<- coffee_summary_long %>%
coffee_summary_wide pivot_wider(names_from = "attribute",
values_from = "mean_score")
ggradar(coffee_summary_wide)
We are going to fix our labels and chanage some parameters on the plot to make it look nicer.
# set our pretty coffee labels
# ggradar plots in alphabetical order so that is how we will label here
<- c("Acidity",
coffee_labels "Aftertaste",
"Aroma",
"Balance",
"Body",
"Clean cup",
"Cupper points",
"Flavor",
"Sweetness",
"Uniformity")
ggradar(coffee_summary_wide,
axis.labels = coffee_labels,
legend.position = "bottom",
axis.label.size = 3,
grid.label.size = 5) +
theme(legend.key = element_rect(fill = NA, color = NA),
plot.title = element_text(size = 16),
legend.text = element_text(size = 12)) +
labs(title = "Difference in average coffee cupper score \nin Arabica and Robusta beans")
Heatmaps π₯β¬οΈπ¦
Install
install.packages("pheatmap")
Load libraries
library(pheatmap)
Plot
pheatmap(mtcars)
pheatmap(mtcars,
scale = "column",
cluster_rows = TRUE) # cluster rows based on similarity
ConplexHeatmap
The package ComplexHeatmap
allows more customized and complicated heatmaps to be produced. If you are interested in making heatmaps, this package is worth to check out.