Lecture 16: Data Visualization

class: center, middle, inverse, title-slide

.title[
# Lecture 16: Data Visualization
]
.author[
### Robin Liu
]
.institute[
### UCSB
]
.date[
### 2022-07-19
]

---

class: inverse, middle, center
# Data Visualization

---
# Data Visualization
Remember the *datasaurus dozen*. Numerical summaries are not enough.

.center[![](Lec16_files/datasaurus.gif)]

---
# Data Visualization
A plot from [Jackman (1980)](https://www.jstor.org/stable/2095134),
a comment on [Hewitt (1977)](https://www.jstor.org/stable/2094750).

.pull-left[
.center[![](Lec16_files/jackman.png)]
]

.pull-right[
Original claim: strong association between voter turnout and income inequality.

Simple scatter plot shows South Africa has *high leverage*.
]

---
# Mapping aesthetics
We have some data in a data frame.

To plot it, we map visual elements (aesthetics) of the plot
to variables in the data frame.

The `ggplot2` library, part of the `tidyverse`, is based on this idea.

Data visualization is an art.
There are no strict guidelines on effective visualization.

---
# Mapping aesthetics
A figure with perceptual problems and excessive aesthetic mappings.

.center[![:scale 60%](Lec16_files/ch-01-chartjunk-life-expectancy.png)]

---
# Mapping aesthetics
Poor choice of visual elements in a plot.

.center[![:scale 50%](Lec16_files/ch-01-excel-3d-column-chart-values.png)]

.center[![:scale 50%](Lec16_files/ch-01-preception-data-1.png)]

Values are mapped to the length of bars, but visual choices
make comparisons difficult.

---
# Mapping aesthetics
### "The best statistical graphic ever drawn"
.center[[![:scale 85%](Lec16_files/minard.png)](https://socviz.co/assets/ch-01-minard.png)]

[Minard: Napoleon's retreat from Moscow](https://sciencenorway.no/blog-blog-from-numbers-to-graphics-statistics/charles-joseph-minards-map-of-napoleons-flawed-russian-campaign-an-ever-current-classic/1618695)

Many variables are displayed:
position and size of army, time advancing and retreating, temperature during retreat.

---
# Mapping aesthetics
We have some data in a data frame.

To plot it, we map visual elements (aesthetics) of the plot
to variables in the data frame.

The `ggplot2` library, part of the `tidyverse`, is based on this idea.

Before introducing `ggplot2`, we look at some examples of plots and their data.

---
# Plot examples

.center[[![:scale 60%](Lec16_files/algebra.png)](https://astralcodexten.substack.com/p/nobody-knows-how-well-homework-works)]

.pull-left[

```
## # A tibble: 4 x 2
##   alg_prog hw_length
##      <dbl> <fct>    
## 1     0.6  0        
## 2     1.1  50       
## 3     1.25 100      
## 4     1.3  150
```
]

.pull-right[
**Bar chart**

- x-axis *mapped* to `hw_length`

- y-axis *mapped* to `alg_prog`
]

---

# Plot examples

.center[[![:scale 60%](Lec16_files/fusion.png)](https://astralcodexten.substack.com/p/your-book-review-the-future-of-fusion)]

.pull-left[

```
## # A tibble: 5 x 3
##    year budget funding_plan
##   <dbl>  <dbl> <fct>       
## 1  1976    0.7 A           
## 2  1976    0.7 B           
## 3  1976    0.7 C           
## 4  1976    0.7 D           
## 5  1977    1   A
```
]

.pull-right[
**Line Plot**
- x-axis *mapped* to `year`
- y-axis *mapped* to `budget`
- color *mapped* to `funding_plan`
]

---
# Plot examples
[Employee Donation](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F416ec862-3759-4e54-87bf-67158501ad96_1710x1562.jpeg)

.center[
![:scale 40%](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F416ec862-3759-4e54-87bf-67158501ad96_1710x1562.jpeg)
]

.pull-left[

```
## # A tibble: 2 x 3
##   employer employee_count biden_pct
##   <chr>             <dbl>     <dbl>
## 1 Google              135        95
## 2 Harvard              13        98
```
]

.pull-right[
** Bubble plot **

- y-axis *mapped* to `biden_pct`
- color *mapped* to `biden_pct`
- size *mapped* to `employee_count`
]

---
class: inverse, middle, center

# Introducing ggplot

---
# ggplot

ggplot examples
- [R Graph Gallery](https://r-graph-gallery.com/ggplot2-package.html)

Resources:
- [Healy - Data Visualization](https://socviz.co/)
- [Wickam - ggplot2](https://ggplot2-book.org/index.html)
- [Wickam - R for Data Science](https://r4ds.had.co.nz/index.html)

---
# Base R vs ggplot
## Redoing the life expectancy plot

.pull-left[![](Lec16_files/ch-01-chartjunk-life-expectancy.png)]
.pull-right[

```r
# library(gapminder)
(gap_avglife <-
   gapminder |>
    filter(year == 2007) |>
    group_by(continent) |>
    summarize(avgLifeExp = mean(lifeExp)))
```

```
## # A tibble: 5 x 2
##   continent avgLifeExp
##   <fct>          <dbl>
## 1 Africa          54.8
## 2 Americas        73.6
## 3 Asia            70.7
## 4 Europe          77.6
## 5 Oceania         80.7
```
]
---
# group_by and summarize

```r
gapminder |>
  filter(year == 2007) |>               # For all observations in 2007,
  group_by(continent) |>                # for each continent...
  summarize(avgLifeExp = mean(lifeExp)) # get the mean life expectancy.
```

```
## # A tibble: 5 x 2
##   continent avgLifeExp
##   <fct>          <dbl>
## 1 Africa          54.8
## 2 Americas        73.6
## 3 Asia            70.7
## 4 Europe          77.6
## 5 Oceania         80.7
```

---
# Base R vs ggplot

```r
par(mar = c(5,8,3,3))
barplot(gap_avglife$avgLifeExp, horiz = T,
        names.arg = c("Africa", "Americas", "Asia", "Europe", "Oceania"), las = "1",
        main = "Average Life Expectancy (2007)", xlab = "Years", ylab = "Continent")
```

`horiz`? `las`? `names.arg`? `mar`? 🤔

The creation process involved seemingly random parameters.
It is missing a *grammar*.

---
# Base R vs ggplot

```r
# library(ggplot2)
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent, fill = continent))
p + geom_col() +
    guides(fill = "none") +
    labs(x = "Years", y = "Continent",
         title = "Average Life Expectancy (2007)",
         caption = "Source: Gapminder.") +
    theme_bw()
```

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent))
p
```

`mapping = aes(x = avgLifeExp, y = continent)`

This defines an **aesthetic mapping** from the data to the visual elements.
The axes reflect this mapping.

The specific visual elements are not decided yet; nothing is plotted.

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent))
* p + geom_col() 
```

We added a **geometric layer** that uses the aesthetic mapping.

`geom_col()` displays the `x` aesthetic as lengths of a bar
and the `y` aesthetic as the category.

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent))
* p + geom_point() 
```

We used a different geometric layer.

`geom_point()` displays the `x` and `y` aesthetics as positions of points.
Note that we used the same mapping!

This is called a *Cleveland dot plot*.

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent, color = continent))
p + geom_col()
```

Adding a new aesthetic mapping. Attempt to color in each bar.

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent, fill = continent))
p + geom_col()
```

Adding a new aesthetic mapping, the continent factor now maps to fill color.

`geom_col` automatically knows how to deal with this mapping.

---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent, fill = continent))
p + geom_col() +
*  guides(fill = "none") 
```

Every aesthetic mapping has a *guide*. Here we set the fill guide to "none",
which removes the legend, since it is redundant.
---
# ggplot

```r
p <- ggplot(gap_avglife, mapping = aes(x = avgLifeExp, y = continent, fill = continent))
p + geom_col() +
    guides(fill = "none") +
    labs(x = "Years", y = "Continent",
         title = "Average Life Expectancy (2007)",
         caption = "Source: Gapminder.")
```

Adjust title and labels with `labs`.

---
# ggplot

Plot can be customized through [many themes](https://r-charts.com/ggplot2/themes/).

---
# ggplot
In week 2 we created this base R plot using `gapminder` data.

Let's recreate this plot in `ggplot`.

```r
plot(lifeExp ~ log(gdpPercap, base = 10), gapminder,
   main = "Life Expectancy and GDP on log scale")
with(gapminder, abline(lm(lifeExp ~ log(gdpPercap, base = 10)), col = "red", lwd = 2))
```

---
# ggplot
<img src="Lec16_files/figure-html/unnamed-chunk-18-1.png" style="display: block; margin: auto;" />

---
# The Layered Grammar of Graphics
.center[![:scale 50%](Lec16_files/gglayers.png)]

There is a learning curve with `ggplot`, but all `ggplot` figures
follow the same *grammar*.

Compare to base R where there are specific parameters for specific functions.

---
# A boxplot
Create the following plot with `ggplot`. You may have to Google (or guess)
what the right geometric layer is.

<div class="countdown" id="timer_62d6d4b4" style="right:0;bottom:0;" data-warnwhen="0">
<code class="countdown-time"><span class="countdown-digits minutes">05</span><span class="countdown-digits colon">:</span><span class="countdown-digits seconds">00</span></code>
</div>
---
class: inverse, center, middle
# Mapping vs Setting Aesthetics

---
# Mapping vs Setting Aesthetics
The x and y positions are mapped to `gdpPercap` and `lifeExp`.

```r
p <- ggplot(gapminder, mapping = aes(x = gdpPercap, y = lifeExp))
p + geom_point() +
    scale_x_log10(labels = scales::label_dollar())
```

---
# Mapping vs Setting Aesthetics
We can map additional variables to aspects of a scatterplot.
For example, we can map the `color` aesthetic to the `continent` variable.

```r
p <- ggplot(gapminder, mapping = aes(x = gdpPercap, y = lifeExp, color = continent))
p + geom_point() +
    scale_x_log10(labels = scales::label_dollar())
```

Note a *guide* was automatically added for this aesthetic.
---
# Mapping vs Setting Aesthetics

```r
p <- ggplot(gapminder, mapping = aes(x = gdpPercap, y = lifeExp, color = "purple"))
p + geom_point() +
    scale_x_log10(labels = scales::label_dollar())
```

What happened?
---
# Mapping vs Setting Aesthetics

Only *mappings* should go in the `aes()` function.
Our tibble does not contain a variable called "purple".

Behind the scenes, `ggplot` created a new categorical variable with "purple" for every row.

```
## # A tibble: 5 x 7
##   country     continent  year lifeExp      pop gdpPercap new_var
##   <fct>       <fct>     <int>   <dbl>    <int>     <dbl> <chr>  
## 1 Afghanistan Asia       1952    28.8  8425333      779. purple 
## 2 Afghanistan Asia       1957    30.3  9240934      821. purple 
## 3 Afghanistan Asia       1962    32.0 10267083      853. purple 
## 4 Afghanistan Asia       1967    34.0 11537966      836. purple 
## 5 Afghanistan Asia       1972    36.1 13079460      740. purple
```

Hence every row with value "purple" is mapped to the same color, which by default is `#F8766D`

In short: `aes()` is for **mappings only**. 
---
# Mapping vs Setting Aesthetics
Setting an attribute outside of `aes()` does what we want.

```r
p <- ggplot(gapminder, mapping = aes(x = gdpPercap, y = lifeExp))
p + geom_point(color = "purple") +
    scale_x_log10(labels = scales::label_dollar())
```

---
# Mapping vs Setting Aesthetics
This can be quite useful.

```r
p <- ggplot(gapminder, mapping = aes(x = gdpPercap, y = lifeExp))
p + geom_point(alpha = 0.3) +
    scale_x_log10(labels = scales::label_dollar())
```

---
# A Polished Plot

<img src="Lec16_files/figure-html/unnamed-chunk-26-1.png" style="display: block; margin: auto;" />
---
# Themes
Many themes are available to make your plots look professional.

Our class will be using default themes:

https://ggplot2.tidyverse.org/reference/ggtheme.html

But know that very fancy themes exist in the `ggthemes` package.

https://yutannihilation.github.io/allYourFigureAreBelongToUs/ggthemes/

---
# Remark
By the end of the class you will be able to identify many `ggplot`s in the wild.

*Exercise:* In the PSTAT hallway in South Hall, there are several posters
displaying professors' research.
Identify the figures that were made with `ggplot`.