03-polish.Rmd

---
title: "Segment 3: Provide Context with Annotations"
description: "Part 3 of the Pearson Live Training Session “Hands–On Data Visualization with ggplot2” for O’Reilly"
author:
  - name: Cédric Scherer
    url: https://cedricscherer.com
output:
  distill::distill_article:
    toc: true
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, warning = FALSE, message = FALSE, 
                      fig.width = 10, fig.height = 6, fig.retina = 2,
                      dev = "ragg_png", res = 1000)
```

```{r preparation}
library(tidyverse)
theme_set(theme_light(base_size = 18))
data <- readr::read_csv("https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/data/crypto_cleaned.csv")
```

## Adjust Colors and Other `aes`thetics

One can use `scale_*()` to change properties of all the **aesthetic dimensions mapped to the data**.

Consequently, there are `scale_*()` functions for all aesthetics such as:

* **positions** via `scale_x_*()` and `scale_y_*()`
* **colors** via `scale_color_*()` and `scale_fill_*()`
* **sizes** via `scale_size_*()` and `scale_radius_*()`
* **shapes** via `scale_shape_*()`  and `scale_linetype_*()`
* **transparency** via `scale_alpha_*()`

The extensions (`*`) can be filled by e.g.:

* `continuous()`, `discrete()`, `reverse()`, `log10()`, `sqrt()`, `date()` for positions
* `continuous()`, `discrete()`, `manual()`, `gradient()`, `gradient2()`, `brewer()` for colors
* `continuous()`, `discrete()`, `manual()`, `ordinal()`, `area()`, `date()` for sizes
* `continuous()`, `discrete()`, `manual()`, `ordinal()` for shapes
* `continuous()`, `discrete()`, `manual()`, `ordinal()`, `date()` for transparency

### Scales: `scale_*()`

Scales are directly connected to aesthetics:

```{r structure-scales}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_x_date() + 
  scale_y_continuous() + 
  scale_color_discrete()
```

All scales come with some general and specific arguments to change the appearance:

```{r structure-scales-adj}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_x_date(
    expand = c(0, 0), ## general
    date_breaks = "4 months", ## date-only 
    date_labels = "%m/%y", ## date only
    name = NULL ## general 
  ) +
  scale_y_continuous() +  
  scale_color_discrete() 
```

```{r structure-scales-adj2}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_x_date(
    expand = c(0, 0), ## general
    date_breaks = "4 months", ## date-only  
    date_labels = "%m/%y", ## date only
    name = NULL ## general 
  ) +
  scale_y_continuous(
    labels = scales::dollar_format(), ## general 
    sec.axis = dup_axis(name = NULL), ## axis only
    name = "Closing Price" ## general 
  ) + 
  scale_color_discrete()
```

```{r structure-scales-adj3}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_x_date(
    expand = c(0, 0), ## general 
    date_breaks = "4 months", ## date-only 
    date_labels = "%m/%y", ## date only 
    name = NULL ## general 
  ) +
  scale_y_continuous(
    labels = scales::dollar_format(), ## general  
    sec.axis = dup_axis(name = NULL), ## axis only 
    name = "Closing Price" ## general 
  ) + 
  scale_color_discrete(
    type = c("#F0B90B", "#4d4d4e", "#810080", "#00aeff"),  ## color only
    name = "Cryptocurrency:" ## general 
  )
```

![Illustration by Allison Horst](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/ah_continuous_discrete.png)

![Color palette types](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/color-pal-types.png)


### Scales: `scale_color_*()`

All colors and fills that are mapped to **categorical** variables can be manipulated with either `scale_color|fill_discrete()` or `scale_color|fill_manual()`.

```{r structure-scales-color-discrete}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_color_discrete(type = c("#F0B90B", "#4d4d4e","#810080", "#00aeff"))
```

```{r structure-scales-color-manual}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_color_manual(values = c("#F0B90B", "#4d4d4e", "#810080", "#00aeff"))
```

Here, you can overwrite the labels of your legend items—no need to manipulate the data itself!

```{r structure-scales-color-manual-labels}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_color_manual(
    values = c("#F0B90B", "#4d4d4e", "#810080", "#00aeff"),
    labels = c("Binance", "Bitcoin", "EOS", "Litecoin"),
    name = "Cryptocurrency:" 
  )
```

![Comparison of continuous color palettes, taken from the {viridis} package reference](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/rainbow_alternatives.png)

![The same collection of palettes for diferent color vision deficiencies](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/rainbow_alternatives_coblis.png)

There are a few and well crafted built-in palettes you can use as well (the defaults are pretty bad).

### Scales: `scale_color_viridis_*()`

```{r structure-scales-color-viridis-default}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_color_viridis_d()
```

```{r structure-scales-color-viridis-turbo}
ggplot(data, aes(x = date, y = close, color = currency)) +
  geom_line() +
  scale_color_viridis_d(
    option = "turbo",
    begin = .1,
    end = .9,
    name = "Cryptocurrency:"
  )
```

![Color palettes contained in the latest version of the `{viridis}` package, taken from the {viridis} package reference](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/viridis-scales.png)

The viridis palettes are more suitable for continuous data:

```{r structure-scales-color-viridis-cont}
ggplot(data, aes(x = date, y = close, color = yday, group = currency)) +
  geom_line() +
  scale_color_viridis_c(
    option = "turbo",
    begin = .1,
    end = .9, 
    name = "Julian Day:"
  )
```

```{r structure-scales-color-viridis-cont-rocket}
ggplot(data, aes(x = date, y = close, color = open, group = currency)) +
  geom_point() +
  scale_color_viridis_c(
    option = "rocket",
    direction = -1,
    end = .9,
    labels = scales::dollar_format(),
    name = "Julian Day:"
  )
```

```{r structure-scales-color-breaks}
ggplot(data, aes(x = date, y = close, color = open, group = currency)) +
  geom_point() +
  scale_color_viridis_c(
    option = "rocket",
    direction = -1, 
    end = .9,
    breaks = seq(50, 300, by = 100),
    labels = scales::dollar_format(), 
    name = "Julian Day:"
  )
```

### Scales: `scale_color_brewer_*()`

Colorbrewer provides color schemes for maps.

```{r brewer-apls}
RColorBrewer::display.brewer.all()
```

```{r structure-scales-color-brewer-default}
ggplot(data, aes(x = date, y = close, color = currency)) +
    geom_line() +
    scale_color_brewer(
        palette = "Set2",
        name = "Cryptocurrency:"
    )
```

---

### Exercise 1

Colorbrewer provides color schemes for <u>maps</u>.

* Download the country data set from `{rnaturalearth}`.
  + see segment 2 if you don't remember how
* Visualize the economy classes (`economy`) as a choropleth map and use one of the categorical colorbrewer palettes.
* Now change the code so the fill colors encode the estimated population (`pop_est`). 
  + What's the problem? How can you fix it?
  + Tip: Have a look at the note and examples on the help page: `?scale_fill_brewer`

---

### Scales: `scale_fill_gradient()`</code>`

You can also build your own **sequential color palettes**:

```{r map-gradient}
sf_world <- rnaturalearth::ne_countries(returnclass = "sf")

ggplot(sf_world) +
  geom_sf(aes(fill = pop_est), color = NA) + 
  scale_fill_gradient(low = "grey90", high = "firebrick4") +
  coord_sf(crs = "+proj=moll")
```

### Scales: `scale_fill_gradient2()`</code>`

You can also build your own **diverging color palettes**:

```{r map-gradient-div}
ggplot(sf_world) +
  geom_sf(aes(fill = as.numeric(factor(income_grp))), color = NA) +
  scale_fill_gradient2(low = "firebrick4", high = "darkcyan", mid = "grey90", name = "Income group:") +
  coord_sf(crs = "+proj=moll")
```

```{r map-gradient-div-midpoint}
ggplot(sf_world) +
  geom_sf(aes(fill = as.numeric(factor(income_grp))), color = NA) +
  scale_fill_gradient2(low = "firebrick4", high = "darkcyan", mid = "grey90", name = "Income group:", 
                       midpoint = 3) +
  coord_sf(crs = "+proj=moll")
```

### Color Palettes in `R`

Several packages offer predefined palettes, e.g.:

*  `{viridis}` for perceptually uniform palettes 
*  `{scico}` for more perceptually uniform palettes 
*  `{rcartocolor}` for map color palettes 
*  `{ggsci}` for scientific journal and sci-fi themed color 
*  `{ggthemes}` for colors of popular software & publishers 
*  `{LaCroixColoR}` for vibrant summery colors 

Check the [collection by Emil Hvitfeldt](https://github.com/EmilHvitfeldt/r-color-palettes/blob/master/type-sorted-palettes.md) for an extensive list of color palettes available in `R`


## A ggplot Object

By the way, you can store ggplots in objects and extend them later:

```{r map-store-object}
g <- ggplot(sf_world) +
  geom_sf(aes(fill = as.numeric(factor(income_grp))), color = NA) +
  scale_fill_gradient2(low = "firebrick4", high = "darkcyan", mid = "grey90", name = "Income group:", 
                       midpoint = 3) +
  coord_sf(crs = "+proj=moll")

class(g)
```


## Legends (Guides)

### Style Legends: `guides()`

The `guides()` function allows to adjust the appearance of each legend:

```{r guide-default}
g +
  guides(fill = guide_colorbar())
```

```{r guide-legend }
g +
  guides(fill = guide_legend())
```

```{r guide-legend-rev }
g +
  guides(fill = guide_legend(reverse = TRUE))
```

```{r guide-legend-height }
g +
  guides(fill = guide_legend(reverse = TRUE, keyheight = unit(2, "lines")))
```

```{r guide-colorbar-height-width }
g +
  guides(fill = guide_colorbar(barheight = unit(12, "lines"),
                               barwidth = unit(1, "lines")))
```

---

## Exercise 2

* Create the following visualization with a viridis palette:
  ![](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/img/exercise-3-2.png)

---

### Style Legends: `theme()`

The position of the legend and some other properties can be changed via `theme()`:

```{r guide-colorbar-theme}
g +
  guides(fill = guide_legend()) +
  theme(legend.position = "top")
```

```{r guide-colorbar-theme-key-size}
g +
  guides(fill = guide_legend()) +
  theme(legend.position = "top", legend.key.height = unit(.5, "lines"),
        legend.key.width = unit(4, "lines"))
```

You can change the position of the legend title and labels in `guides()`:

```{r guide-colorbar-theme-legendtitlepos}
g +
  guides(fill = guide_legend(title.position = "top", title.hjust = .5, label.position = "bottom")) +
  theme(legend.position = "top", legend.key.height = unit(.5, "lines"),
        legend.key.width = unit(4, "lines")) 
```

## Template and Custom Theming

### Themes: `theme_*()`

You have already seen built-in themes in segment 1:

```{r themes}
gg <- ggplot(data, aes(x = date, y = close, color = currency)) +
       geom_line(size = 1.5)

gg +
  theme_grey() +
  ggtitle("theme_grey()  or  theme_gray()") 

gg +
  theme_classic() +
  ggtitle("theme_classic()") 

gg +
  theme_bw() +
  ggtitle("theme_bw()") 

gg +
  theme_light() +
  ggtitle("theme_light()")

gg +
  theme_minimal() +
  ggtitle("theme_minimal()") 

gg +
  theme_dark() +
  ggtitle("theme_dark()") 

gg +
  theme_linedraw() +
  ggtitle("theme_linedraw()")

gg +
  theme_void() +
  ggtitle("theme_void()") 
```

You can either add theme changes to each plot or globally for all plots:


```{r theme-local}
ggplot(data, aes(x = date, y = close)) +
  geom_line(aes(color = currency)) + 
  theme_classic(base_size = 16)
```

```{r theme-global}
theme_set(theme_classic(base_size = 16))
ggplot(data, aes(x = date, y = close)) +
  geom_line(aes(color = currency))
```

```{r reset-theme, echo=FALSE}
theme_set(theme_light(base_size = 18))
```

```{r theme-function}
theme_grey
```

### `theme()` Arguments

There are many elements you can customize. You can either group them by their type or by their category:

* element types: 
  - `text` → all labels, axis text, legend title and text
  - `line` → axis lines, ticks, grid lines
  - `rect` → plot area, panel area, legend and legend keys, facets


* element category:
  - `axis.*` → titles, text, ticks, lines
  - `legend.*` → background, margin, spacing, keys, text, title, position, direction, box
  - `panel.*` → background, border, margin, spacing, grid (major and minor)
  - `plot.*` → background, title, subtitle, caption, tag, margin
  - `strip.*` → background, placement, text

You can directly alter the appearance by adding `theme()` to a ggplot.


#### Rectangular Elements via `element_rect()`

![Theme components that are `element_rect` (Part 1)](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/element-rect-1.png)

![Theme components that are `element_rect` (Part 2)](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/element-rect-2.png)


```{r element-rect-example}
g +
  theme(plot.background = element_rect(
    color = "black",
    fill = "orange",
    size = 2,
    linetype = "dotted"
  ))
```


#### Line Elements via `element_line()`

![Theme components that are `element_line`](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/element-line.png)

```{r element-line-example}
g +
  theme(panel.grid = element_line(
    color = "black",
    size = 3,
    linetype = "dashed",
    lineend = "square", # round, butt
    arrow = arrow(angle = 30, length = unit(0.25, "inches"))
  ))
```

#### Text Elements via `element_text()`

![Theme components that are `element_text`](https://raw.githubusercontent.com/z3tt/hands-on-ggplot2/main/slides/img/element-text.png)


```{r element-text-example}
g +
  ggtitle("My New\nTitle") +
  theme(plot.title = element_text(
    #family = "Roboto", ## you need {systemfonts} for this
    face = "bold", ## plain, italic, bolditalic
    size = 24,
    color = "firebrick",
    lineheight = .7,
    angle = 180,
    hjust = .5,
    vjust = .0,
    margin = margin(
      10, ## t (top)
      0,  ## r (right)
      30, ## b (bottom)
      0   ## l (left)
    )
  ))
```

---

## Exercise 3

* Have a closer look at the codes of `theme_grey`/`theme_gray` and one of the other themes. What's the difference? 
* Create a theme that is as ugly as possible!  
  or:  
  Create a theme that you can use ass personal theme.  
* Apply it to a plot.

---

## Resources

* [Modify theme components](https://ggplot2.tidyverse.org/reference/theme.html) on the ggplot reference page

* Chapter 19 [Themes](https://ggplot2-book.org/polishing.html) of the “ggplot2” book by Hadley Wickham et al.

* [“Creating and Using Custom ggplot2 Themes”](https://themockup.blog/posts/2020-12-26-creating-and-using-custom-ggplot2-themes/), blog post by Thomas Mock

* [“Themes to Improve Your ggplot Figures”](https://rfortherestofus.com/2019/08/themes-to-improve-your-ggplot-figures/) from [R for the Rest of Us](https://rfortherestofus.com/) with a collection of additional themes

* [“A `{ggplot2}` Tutorial for Beautiful Plotting in R”](https://www.cedricscherer.com/2019/08/05/a-ggplot2-tutorial-for-beautiful-plotting-in-r/), my extensive "how to"-tutorial

---

## Session Info

<details><summary>Expand for details</summary>

```{r sessionInfo, echo = F}
Sys.time()
git2r::repository()
sessionInfo()
```

</details>