Hands-on Introduction to R 2023

1. Introduction: what is data analysis and R basics
2. Data visualization and reporting
3. Data transformation 1: extract, summarize
4. Data transformation 2: join, pivot
5. Data cleansing and conversion: numbers, text
6. Data input and interpretation
7. Statistical modeling 1: probability distribution, likelihood
8. Statistical modeling 2: linear regression

Outline of data analysis

1. Setup computer environment
2. Get and read input data
3. Exploratory data analysis
   • Preparation (harder than it seems) 👈 lecture #3–5
   • Visualization, generating hypotheses (fun!) 👈 today
   • Statistical analysis, testing hypotheses
4. Report

Visualization helps overviewing data structure

Reduction and reorganization of information → intuitive understanding

The larger carat, the higher price.
The slope seems to differ by clarity.

Never trust summary statics alone

Interesting relationships may be overlooked without visualization.

Visualization requires an innate sense of beauty?

NO!

To some extent,
it is about knowledge and skills.

You can improve it
once you know the basic design rules.

Today’s lesson: data visualization and reporting with R

✅ Outline of data analysis: visualization is important

⬜ Draw graphs in a consistent manner with ggplot2

⬜ Embed R code and results into a report with Quarto

iris: an example dataset

4 numeric and 1 factor vectors with the length of 150:

print(iris)

    Sepal.Length Sepal.Width Petal.Length Petal.Width   Species
  1          5.1         3.5          1.4         0.2    setosa
  2          4.9         3.0          1.4         0.2    setosa
  3          4.7         3.2          1.3         0.2    setosa
  4          4.6         3.1          1.5         0.2    setosa
 --                                                            
147          6.3         2.5          5.0         1.9 virginica
148          6.5         3.0          5.2         2.0 virginica
149          6.2         3.4          5.4         2.3 virginica
150          5.9         3.0          5.1         1.8 virginica

Base R graphics

inconsistent and difficult to customize details.

boxplot(Petal.Width ~ Species, data = iris)
plot(iris$Sepal.Length, iris$Sepal.Width)
hist(iris$Petal.Length)

Any useful package to draw beautiful graphs easily?

ggplot2: a tidyverse member in charge of graphics

• can draw many types of graphs in a rational and consistent manner.
• based on “The Grammar of Graphics” (Wilkinson 2005).

ggplot2: a tidyverse member in charge of graphics

• can draw many types of graphs in a rational and consistent manner.
• based on “The Grammar of Graphics” (Wilkinson 2005).

Learn ggplot2 without knowing base R graphics

No problem. Underlying systems are fundamentally different:

Basic usage: join layers with + operator

Basic usage: join layers with + operator

ggplot(data = diamonds)             # prepare canvas with diamonds data
# aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
# geom_point() +                    # draw scatter plot
# facet_wrap(vars(clarity)) +       # split panels according to clarity
# stat_smooth(method = lm) +        # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price)         # map (carat, price) to (x, y) axes
# geom_point() +                    # draw scatter plot
# facet_wrap(vars(clarity)) +       # split panels according to clarity
# stat_smooth(method = lm) +        # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point()                      # draw scatter plot
# facet_wrap(vars(clarity)) +       # split panels according to clarity
# stat_smooth(method = lm) +        # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point() +                    # draw scatter plot
  facet_wrap(vars(clarity))         # split panels according to clarity
# stat_smooth(method = lm) +        # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point() +                    # draw scatter plot
  facet_wrap(vars(clarity)) +       # split panels according to clarity
  stat_smooth(method = lm)          # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point() +                    # draw scatter plot
  facet_wrap(vars(clarity)) +       # split panels according to clarity
  stat_smooth(method = lm) +        # add linear regression
  coord_cartesian(ylim = c(0, 2e4))   # adjust y-axis range
# theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point() +                    # draw scatter plot
  facet_wrap(vars(clarity)) +       # split panels according to clarity
  stat_smooth(method = lm) +        # add linear regression
  coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
  theme_classic(base_size = 20)     # adjust non-data components

Basic usage: join layers with + operator

ggplot(data = diamonds) +           # prepare canvas with diamonds data
  aes(x = carat, y = price) +       # map (carat, price) to (x, y) axes
  geom_point() +                    # draw scatter plot
# facet_wrap(vars(clarity)) +       # split panels according to clarity
# stat_smooth(method = lm) +        # add linear regression
# coord_cartesian(ylim = c(0, 2e4)) + # adjust y-axis range
  theme_classic(base_size = 20)     # adjust non-data components

Intermediate states can be saved as objects

p1 = ggplot(data = diamonds)
p2 = p1 + aes(x = carat, y = price)
p3 = p2 + geom_point()
p4 = p3 + facet_wrap(vars(clarity))
print(p3)

We are going to use this p3 later.

Let’s try ggplot

using mpg, a dataset of fuel economy:

    manufacturer  model displ year cyl      trans drv cty hwy fl   class
  1         audi     a4   1.8 1999   4   auto(l5)   f  18  29  p compact
  2         audi     a4   1.8 1999   4 manual(m5)   f  21  29  p compact
 --                                                                     
233   volkswagen passat   2.8 1999   6 manual(m5)   f  18  26  p midsize
234   volkswagen passat   3.6 2008   6   auto(s6)   f  17  26  p midsize

🔰 Scatter-plot the relationship of displ and cty
(engine displacement in litres and city miles per gallon)

Common errors

No function is named ggplot2:

> ggplot2(diamonds)
Error in ggplot2(diamonds) : could not find function "ggplot2"

ggplot2 is a package, which has a function named ggplot.
OK, but still R cannot find the function:

> ggplot(diamonds)
Error in ggplot(diamonds) : could not find function "ggplot"

Don’t forget to load the package every time you launch R/RStudio:

library(conflicted) # charm for safe coding
library(tidyverse)  # load packages including ggplot2
ggplot(diamonds)    # OK!

ggplot() requires “tidy data”

• Each column is a variable.
• Each row is an observation.
• Each cell is a single value.

print(diamonds)

      carat       cut color clarity depth table price    x    y    z
    1  0.23     Ideal     E     SI2  61.5    55   326 3.95 3.98 2.43
    2  0.21   Premium     E     SI1  59.8    61   326 3.89 3.84 2.31
    3  0.23      Good     E     VS1  56.9    65   327 4.05 4.07 2.31
    4  0.29   Premium     I     VS2  62.4    58   334 4.20 4.23 2.63
   --                                                               
53937  0.72      Good     D     SI1  63.1    55  2757 5.69 5.75 3.61
53938  0.70 Very Good     D     SI1  62.8    60  2757 5.66 5.68 3.56
53939  0.86   Premium     H     SI2  61.0    58  2757 6.15 6.12 3.74
53940  0.75     Ideal     D     SI2  62.2    55  2757 5.83 5.87 3.64

Map variables to aesthetics

by specifying column names in aes():

ggplot(diamonds) +
  aes(x = carat, y = price) +
  geom_point(mapping = aes(color = clarity, size = cut))

Set constant/independent aesthetics

by specifying values out of aes():

ggplot(diamonds) +
  aes(x = carat, y = price) +
  geom_point(color = "darkorange", size = 6, alpha = 0.4)

Outer aes() propagates to all geom_*()

ggplot(diamonds) +
  aes(x = carat, y = price) +
  geom_point(aes(color = clarity)) +
  geom_line()             # NO color
ggplot(diamonds) +
  aes(x = carat, y = price, color = clarity) +
  geom_point() +          # color
  geom_line()             # color

aesthetics list

• Color related
  • color: outline color of points, lines, and characters
  • fill: fill color of shapes
  • alpha: opacity (0 = transparent, 1 = opaque)
• Differentiation related
  • size, shape: of points and shapes
  • linewidth, linetype
• Grouping
  • group: to separate lines and polygons
• Position related
  • x, y, xmin, xmax, ymin, ymax, xend, yend

color for points, lines, characters; fill for areas

alpha for opacity.

ggplot(diamonds) +
  aes(cut, carat) +
  geom_boxplot(color = "royalblue", fill = "gold", alpha = 0.5, linewidth = 2)

Practice to change colors

using mpg, a dataset of fuel economy:

    manufacturer  model displ year cyl      trans drv cty hwy fl   class
  1         audi     a4   1.8 1999   4   auto(l5)   f  18  29  p compact
  2         audi     a4   1.8 1999   4 manual(m5)   f  21  29  p compact
 --                                                                     
233   volkswagen passat   2.8 1999   6 manual(m5)   f  18  26  p midsize
234   volkswagen passat   3.6 2008   6   auto(s6)   f  17  26  p midsize

🔰 Draw scatter plot of displ and cty with blue points.
🔰 Color-code points according to drv and cyl (drive train and # of cylinders).

Diversity in color vision

The previous graph with 3 colors, red green blue, is no problem for many people.
But for ~5% people, it looks like two-color: red green blue or red green blue.

You can simulate color vision with a Mac/iOS app, Sim Daltonism.
Color Oracle is a Windows app.

Some palettes are designed with consideration

Sequential palette:

Diverging palette:

Qualitative (categorical, discrete) palette:

Change color palette with scale_color_*()

viridis and ColorBrewer are included in ggplot2.
Find names from the link above, and specify it with option = or palette =.

ggplot(diamonds) + aes(carat, price) +
  geom_point(mapping = aes(color = clarity)) +
  scale_color_viridis_d(option = "inferno")
# scale_color_brewer(palette = "YlGnBu")

Distinguish continuous and discrete variables

and choose a scale_* function accordingly, or you will get
Error: Continuous value supplied to discrete scale

ggplot(diamonds) + aes(carat, price) +
  geom_point(mapping = aes(color = price)) +
  scale_color_viridis_c(option = "inferno")
# scale_color_distiller(palette = "YlGnBu")

• discrete: scale_color_viridis_d(), scale_color_brewer()
• continuous: scale_color_viridis_c(), scale_color_distiller()
• binned: scale_color_viridis_b(), scale_color_fermenter()

Useful palettes other than viridis and brewer

Try built-in palette.colors() and colorspace package.

okabe_ito = palette.colors(9L, "Okabe-Ito")
ggplot(mpg) +
  aes(x = displ, y = cty) +
  geom_point(aes(color = drv), size = 4, alpha = 0.66) +
  scale_color_discrete(type = unname(okabe_ito)[-1])
# scale_color_discrete(type = palette.colors(8L, "R4")[-1])
# colorspace::scale_colour_discrete_divergingx("Zissou 1")

These palettes are carefully designed.

Global options to omit scale_color_*

An example setting to use viridis and Okabe-ito by default:

grDevices::palette("Okabe-Ito")
options(
  ggplot2.continuous.colour = "viridis",
  ggplot2.continuous.fill = "viridis",
  ggplot2.discrete.colour = grDevices::palette()[-1],
  ggplot2.discrete.fill = grDevices::palette()[-1]
)

Settings by options() are effective until an R session ends.

Facet: splitting a plot into a sequence of panels

One of the biggest benefits of using ggplot!

p3 + facet_wrap(vars(clarity), ncol = 4L)

Facet: splitting a plot into a 2d grid of panels

One of the biggest benefits of using ggplot!

p3 + facet_grid(vars(clarity), vars(cut))

Overviewing multivariate data with “aes” + “facet”

Practice of faceting

using mpg, a dataset of fuel economy:

    manufacturer  model displ year cyl      trans drv cty hwy fl   class
  1         audi     a4   1.8 1999   4   auto(l5)   f  18  29  p compact
  2         audi     a4   1.8 1999   4 manual(m5)   f  21  29  p compact
 --                                                                     
233   volkswagen passat   2.8 1999   6 manual(m5)   f  18  26  p midsize
234   volkswagen passat   3.6 2008   6   auto(s6)   f  17  26  p midsize

🔰 Try faceting by drv, cyl, and both.

Modify coordinate axes with scale_*, coord_*

ggplot(diamonds) + aes(carat, price) + geom_point(alpha = 0.25) +
  scale_x_log10() +
  scale_y_log10(breaks = c(1, 2, 5, 10) * 1000) +
  coord_cartesian(xlim = c(0.1, 10), ylim = c(800, 12000)) +
  labs(title = "Diamonds", x = "Size (carat)", y = "Price (USD)")

Customize non-data components with theme

Choose a complete theme_*(), and adjust each element with theme().

p3 + theme_bw(base_size = 18) + theme(
  panel.background = element_rect(fill = "khaki"),      # box
  panel.grid       = element_line(color = "royalblue"),
  axis.title.x     = element_text(size = 32),
  axis.text.y      = element_blank()                    # remove
)

Basic usage: join layers with + operator

Combine separate plots into one as in papers

with the help of another package such as cowplot and patchwork:

pAB = cowplot::plot_grid(p3, p3, labels = c("A", "B"), nrow = 1L)
cowplot::plot_grid(pAB, p3, labels = c("", "C"), ncol = 1L)

Save plots in a reproducible way

RStudio’s “Export” button is affected by the window size.
Adjust and fix the size with ggsave().

# 7 inch x 300 dpi = 2100 px square (default)
ggsave("dia1.png", p3) # width = 7, height = 7, dpi = 300
# 4      x 300     = 1200  (Zoom in x7/4)
ggsave("dia2.png", p3, width = 4, height = 4) # dpi = 300
# 2      x 600     = 1200  (Zoom in x7/2)
ggsave("dia3.png", p3, width = 2, height = 2, dpi = 600)
# 4      x 300     = 1200  (Magnify characters in theme components)
ggsave("dia4.png", p3 + theme_bw(base_size = 22), width = 4, height = 4)

Show multi-byte characters properly, not tofu ◻◻

Preferences → General → Graphics → Backend: AGG

(Using non-ASCII characters should be avoided in the first place…)

What kind of geom_*() are available?

See the official document.

Final code can be overwhelmingly long…

Yes. But you can validate and reuse it later.

set.seed(1)
p = ggplot(diamonds) +
  aes(x = cut, y = price) +
  geom_jitter(aes(color = cut), height = 0, width = 0.2, alpha = 0.1, stroke = 0) +
  geom_boxplot(fill = NA, outlier.shape = NA) +
  scale_color_viridis_d(option = "plasma") +
  facet_wrap(vars(clarity)) +
  coord_flip(xlim = c(0.5, 5.5), ylim = c(0, 20000), expand = FALSE) +
  labs(title = "Diamonds", x = "Cut", y = "Price (USD)") +
  theme_bw(base_size = 20) +
  theme(legend.position = "none",
        axis.ticks = element_blank(),
        panel.grid.major.y = element_blank(),
        panel.spacing.x = grid::unit(3, "lines"),
        plot.margin = grid::unit(c(1, 2, 0.5, 0.5), "lines"))
print(p)
ggsave("diamonds-cut-price.png", p, width = 12, height = 9)

Advanced usage

ggplot2 extensions

gganimate: Animation

ggrepel: Repel overlapping text labels away

ggraph: Networks

ggtree: Phylogenetic trees

ggpubr: Academic publication

ggplot2 does not support 3D plotting.

Really need 3D? Aes and facet will do in most cases.

Try other packages: rgl, plotly, rayshader, etc.

🔰 Today’s challenge 1: reproduction

Draw the same figures as follows:

Take a close look at the details 👀

Today’s lesson: data visualization and reporting with R

✅ Outline of data analysis: visualization is important

✅ Draw graphs in a consistent manner with ggplot2

• Aesthetic mapping and faceting play key roles
• Diversity in color vision
• Save images in a reproducible way

⬜ Embed R code and results into a report with Quarto

If you want to see the results of an R script

Execute lines in the script again — cumbersome

See the image files written by ggsave() — which is from which code?

→ Any format to take a glace at code and its result at the same time?

3 * 14
ggplot(mpg) + aes(displ, hwy) + geom_point(aes(color = drv))

[1] 42

Quarto Document

You can unify management of programming and reporting.

• Create a text file (.qmd) with body text and R code.
• Translate it to a rich format such as HTML and PDF
  • Figures and tables can be embedded as well as R code.

Quarto Markdown (.qmd)

A subspecies of Markdown; .Rmd is almost the same.

Can execute R code and embed the results.

Markdown (.md)

The most popular lightweight markup language.

There are some different flavors. Quarto uses Pandoc Markdown.

(See the gallery for examples.)

Markup languages

Set of rules to express document structures and layouts.
e.g., HTML+CSS, XML, LaTeX

<h3>Heading level 3</h3>
<p>This is a paragraph.
<em>emphasis (italic)</em>,
<strong>strong (bold)</strong>,
<a href="https://www.lifesci.tohoku.ac.jp/">link</a>, etc.
</p>

Expressive and powerful, but too complex for humans to read/write.

(Try viewing the source code of your favorite websites.)

Lightweight markup languages

Markup languages that are easy for humans to read and write.
e.g., Markdown, reStructuredText, various wiki notations

### Heading level 3

This is a paragraph.
*emphasis (italic)*,
**strong (bold)**,
[link](https://www.lifesci.tohoku.ac.jp/), etc.

You already have Quarto environment

• R (≥ 4.3.2): “Latest – 0.1” is acceptable.
• RStudio (≥ 2023.09.1+494): including Quarto CLI
• tidyverse (≥ 2.0.0): installs the following two:
  • rmarkdown (≥ 2.25)
  • knitr (≥ 1.45)

(The versions shown above are not requirement, but my current values.)

• Quarto CLI: if you really want the bleeding edge version.
• install.packages("quarto"): unnecessary for most people; just provides R functions to call Quarto CLI.
• pandoc: installed along with Quarto CLI. not sure if the manually installed version is used by RStudio+Quarto.

Try writing and translating Markdown

1. RStudio > New File > Markdown File
2. Search “markdown syntax” and write something. Include the following elements at least:
   • Heading level 1, 2, 3
   • Code blocks, inline code
   • Bullet lists (with and without item number)
3. Push Preview button and check the result.

Try writing a Quarto Document

RStudio > New File > Quarto Document…
Select “Document” and “HTML”; fill “Title” and “Author”; create.

New features in qmd, not in md

Header (front matter)

Surrounded by --- at the top. Metadata of a whole document.

Options differ depending on output formats. e.g., html

R code chunks

Code blocks that start like ```{r}.

The result of the code is also embedded into the product.

Many options。e.g.,

• echo: false: Hide code, but execute and show its results.
• eval: false: Show code without execution.
• include: false: Execute code, but hide it and its results.
• fig.width: 7, fig.height: 7: Adjust image size.

Don’t care about the details for now.

Try translating qmd to HTML

qmd on the left is a source code. HTML on the right is its product.

Try translating qmd to HTML

1. Save a source code with commands
   e.g., report.qmd
2. Push ⚙️ button, and select “Preview in Viewer Pane”
3. Push →Render button.
   • Embedded R code is executed.
   • Markdown with the results is generated.
   • Markdown is translated to HTML, e.g., report.html
   • Preview is shown on the right automatically.
4. Repeat editing → saving → “Render” to make a satisfactory report.

An example of reports

Create qmd on the left; push “Render” and submit HTML on the right.

参考

R for Data Science — Hadley Wickham et al.

https://r4ds.hadley.nz

Paperback

日本語版書籍(Rではじめるデータサイエンス)

Older versions

「Rにやらせて楽しよう — データの可視化と下ごしらえ」 岩嵜航 2018

「Rを用いたデータ解析の基礎と応用」石川由希 2019 名古屋大学

「Rによるデータ前処理実習」 岩嵜航 2022 東京医科歯科大

「Rを用いたデータ解析の基礎と応用」 石川由希 2023 名古屋大学

ggplot2 official document

https://ggplot2.tidyverse.org/