Data Visualization

class: center, middle, inverse, title-slide

# Data Visualization
## The grammar of graphics and ggplot2
### June 8th, 2022

---

## Do these [datasets](http://www.thefunctionalart.com/2016/08/download-datasaurus-never-trust-summary.html) have anything in common?

.center[![](https://d2f99xq7vri1nk.cloudfront.net/DataDino-600x455.gif)]

---

## __"Graphics reveal data"__- [Edward Tufte](https://www.edwardtufte.com/tufte/)

.center[![](https://www.researchgate.net/profile/Arch_Woodside2/publication/285672900/figure/fig4/AS:305089983074309@1449750528742/Anscombes-quartet-of-different-XY-plots-of-four-data-sets-having-identical-averages.png)]

__Always visualize your data__ before analyzing / modeling it

---

## Florence Nightingale's rose diagram

---

---

## Quick review of yesterday's material

.pull-left[

We are exploring MLB hitting data from [`Lahman`](https://cran.r-project.org/web/packages/Lahman/index.html) using the [`tidyverse`](https://www.tidyverse.org/)

```r
library(tidyverse)
library(Lahman) 
Batting <- as_tibble(Batting) 
year_batting_summary <- Batting %>%
  filter(lgID %in% c("AL", "NL")) %>%
  group_by(yearID) %>%
  summarize_at(vars(H, HR, SO, BB, AB),
               sum, na.rm = TRUE) %>%
  mutate(batting_avg = H / AB)
```

__But how do we make data visualizations in `R`?__

What steps do we take to make this figure?
]

.pull-right[

]

---

### [The Grammar of Graphics](https://www.amazon.com/gp/product/0387245448/ref=as_li_ss_tl?ie=UTF8&camp=1789&creative=390957&creativeASIN=0387245448&linkCode=as2&tag=civilstatis-20)
Principled data visualization framework introduced by [Leland Wilkinson](https://www.cs.uic.edu/~wilkinson/)

- __start with the raw data__ -> visualize transformations, summaries, etc. of the data

[Hadley Wickham](http://hadley.nz/) expanded upon this foundation with a [layered](http://vita.had.co.nz/papers/layered-grammar.pdf) `R` implementation via [`ggplot2`](https://ggplot2.tidyverse.org/)

--
1. __`data`__ - one or more datasets (in tidy tabular format)

--
2. __`geom`__ - one or more geometric objects to visually represent the data (e.g. points, lines, bars, etc.)

--
3. __`aes`__ - mappings of columns / variables to visual properties (i.e. _aesthetics_) of the geometric objects

--
4. __`scale`__ - one scale for each variable displayed (e.g. axis limits, log scale, colors, etc.)

--
5. __`facet`__ - similar subplots (i.e. _facets_) for subsets of the same data using a categorical variable

--
6. __`stat`__ - statistical transformations and summaries (e.g. identity, count, smooth, quantile, etc.)

--
7. __`coord`__ - one or more coordinate systems (e.g. __cartesian__, polar, map projection)

--
8. __`labs`__ - labels/guides for each variable and other parts of the plot (e.g. title, subtitle, caption, etc.)

--
9. __`theme`__ - customization of plot layout (e.g. text size, alignment, legend position, etc.)

---

## Start with the data...

.pull-left[

```r
*ggplot(data = year_batting_summary)
```

__or equivalently using the `%>%`__

```r
year_batting_summary %>%
  ggplot()
```

__but nothing is displayed!__

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-3-1.png" width="504" />
]

---

## What variables and geometric object?

.pull-left[

```r
year_batting_summary %>%
  ggplot() + 
* geom_point(aes(x = yearID, y = HR))
```

- adding (`+`) a geometric layer of points to the plot

- map `yearID` to the x-axis and `HR` to the y-axis via `aes()`

- implicitly using `coord_cartesian()`

```r
year_batting_summary %>%
  ggplot() + 
  geom_point(aes(x = yearID, y = HR)) +
* coord_cartesian()
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-4-1.png" width="504" />
]

---

## Can we add another geometric layer?

.pull-left[

```r
year_batting_summary %>%
  ggplot() + 
  geom_point(aes(x = yearID, y = HR)) +
* geom_line(aes(x = yearID, y = HR))
```

- adding (`+`) a line geometric layer

- Include mappings shared across geometric layers inside `ggplot()`

```r
year_batting_summary %>%
* ggplot(aes(x = yearID, y = HR)) +
  geom_point() +
  geom_line()
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-5-1.png" width="504" />
]

---

## What about the scales?

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() +
* scale_x_continuous() +
* scale_y_continuous()
```

- `yearID` and `HR` are continuous variables, resulting in continuous scales by default
]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-6-1.png" width="504" />
]

---

## We can customize the scale limits

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() +
* scale_x_continuous(limits = c(2000, 2018))
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-7-1.png" width="504" />
]

---

## We can customize the label breaks

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() +
* scale_y_continuous(breaks = seq(0, 6000,
*                                 by = 1000))
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-8-1.png" width="504" />
]

---

## We can use different scales!

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() + 
* scale_x_reverse() +
* scale_y_log10()
```

__You can easily adjust variable scales without modifying the columns directly__

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-9-1.png" width="504" />
]

---

## Add a statistical summary?

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() +
* stat_smooth()
```

- Smoothing regression summary (will cover later) using `yearID` and `HR`

- Geometric layers implicitly use a default statistical summary

- Technically we are already using `geom_point(stat = "identity")`

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point() +
  geom_line() +
* geom_smooth()
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-10-1.png" width="504" />
]

---

## Map additional variables?

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR,
*            color = SO,
*            size = BB)) +
  geom_point() +
  geom_line()
```

- `HR`, `SO`, and `BB` are all displayed!

- `color` and `size` are being shared across layers

- This is a bit odd to look at...

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-11-1.png" width="504" />
]

---

## Customize mappings by layer

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
* geom_point(aes(color = SO,
*                size = BB)) +
  geom_line()
```

- Now mapping `SO` and `BB` to `color` and `size` of __only__ the point layer
]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-12-1.png" width="504" />
]

---

## Can change aesthetics without mapping variables

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point(aes(color = SO, 
                 size = BB)) + 
* geom_line(color = "darkred",
*           linetype = "dashed")
```

- Set manual values to the `color` and `linetype` of the line layer
]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-13-1.png" width="504" />
]

---

## Remember: one scale for each mapped variable

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point(aes(color = SO, 
                 size = BB)) + 
  geom_line(color = "darkred",
            linetype = "dashed") +
* scale_color_gradient(low = "darkblue",
*                      high = "darkorange") +
* scale_size_continuous(breaks =
*                         seq(0, 20000,
*                             2500))
```

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-14-1.png" width="504" />
]

---

## You MUST label your plots!

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point(aes(color = SO, 
                 size = BB)) + 
  geom_line(color = "darkred",
            linetype = "dashed") +
  scale_color_gradient(low = "darkblue", 
                       high = "darkorange") + 
* labs(x = "Year", y = "Homeruns",
*      color = "Strikeouts",
*      size = "Walks",
*      title = "The rise of MLB's three true outcomes",
*      caption = "Data courtesy of Lahman")
```

- Each mapped aesthetic can be labelled (_what happened to the legend order?_)

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-15-1.png" width="504" />
]

---

## Custom theme

.pull-left[

```r
year_batting_summary %>%
  ggplot(aes(x = yearID, y = HR)) + 
  geom_point(aes(color = SO, 
                 size = BB)) + 
  geom_line(color = "darkred",
            linetype = "dashed") +
  scale_color_gradient(low = "darkblue", 
                       high = "darkorange") + 
  labs(x = "Year", y = "Homeruns",
       color = "Strikeouts", 
       size = "Walks", 
       title = "The rise of MLB's three true outcomes",
       caption = "Data courtesy of Lahman") +
* theme_bw() +
* theme(legend.position = "bottom",
*       plot.title = element_text(size = 15),
*       axis.title = element_text(size = 10))
```

- `theme_bw()` is a popular default, but check out [`ggthemes`](https://jrnold.github.io/ggthemes/index.html) for more, ex: `theme_fivethirtyeight()`

]
.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-16-1.png" width="504" />
]

---

## A lesson about data visualization...

.center[![](https://media1.tenor.com/images/02e615bbff46a6f4ed05735360532142/tenor.gif?itemid=4683433)]

--
__Simpler is better__ - instead create three separate plots for `HR`, `SO`, and `BB` with each mapped to `y`

--
_How do we do this without repeating the same code three times?_

---

## Pivot the data!

__Remember__: we need the data in tidy format

Within the `tidyverse`, the [`tidyr`](https://tidyr.tidyverse.org/index.html) package has functions that allow us to easily _realign_ our data

- [`pivot_longer`](https://tidyr.tidyverse.org/reference/pivot_longer.html) - gathers information spread out across variables, increase `nrow()` but decrease `ncol()`

- [`pivot_wider`](https://tidyr.tidyverse.org/reference/pivot_wider.html) - spreads information out  from observations, decrease `nrow()` but increase `ncol()`

.pull-left[

```r
year_batting_summary %>%
  select(yearID, HR, SO, BB) %>%
  rename(HRs = HR, Strikeouts = SO,
         Walks = BB) %>%
* pivot_longer(HRs:Walks,
*              names_to = "stat",
*              values_to = "value")
```

We have now created a new variable `stat`

]
.pull-right[

```
## # A tibble: 6 x 3
##   yearID stat       value
##    <int> <chr>      <int>
## 1   1876 HRs           40
## 2   1876 Strikeouts   589
## 3   1876 Walks        336
## 4   1877 HRs           24
## 5   1877 Strikeouts   726
## 6   1877 Walks        345
```
]

---

## Use facets to create subplots with categorical variables

.pull-left[

```r
year_batting_summary %>%
  select(yearID, HR, SO, BB) %>%
  rename(HRs = HR, Strikeouts = SO,
         Walks = BB) %>%
  pivot_longer(HRs:Walks,
               names_to = "stat", 
               values_to = "value") %>%
  ggplot(aes(x = yearID, y = value)) +
  geom_line(color = "darkblue") +
  geom_point(color = "darkblue") +
* facet_wrap(~ stat,
*            scales = "free_y", ncol = 1) +
  labs(x = "Year", y = "Total of statistic",
       title = "The rise of MLB's three true outcomes",
       caption = "Data courtesy of Lahman") +
  theme_bw() +
  theme(strip.background = element_blank())
```

]

.pull-right[
<img src="02-Ggplot-intro_files/figure-html/unnamed-chunk-18-1.png" width="504" />
]