09_ggplot2.Rmd

---
title: "09_ggplot2"
output: html_document
---

class: middle, center, inverse
layout: false

# 4.7 `ggplot2`:<br><br>Create Elegant Data Visualisations<br>Using the Grammar of Graphics

---

background-image: url(https://raw.githubusercontent.com/tidyverse/ggplot2/master/man/figures/logo.png)
background-position: 97.5% 2.5%
background-size: 7.5%
layout: true

---

## 4.7 `ggplot2`: Elegant Data Visualisations

`ggplot2` is Hadley Wickham's [reimplementation](https://www.tandfonline.com/doi/abs/10.1198/jcgs.2009.07098) of the 2005 published *The Grammar of Graphics* by Leland Wilkinson. It provides a large amount of functions for generating high-quality graphs in a layer-based fashion and has even sparked a whole ecosystem of 'gg'-style visualization packages.

<br>

```{r, echo=F, out.width='75%', fig.align='center'}
knitr::include_graphics("./img/grammar-of-graphic-layers.png")
```

.center[
*Src: [towardsdatascience](https://towardsdatascience.com/a-comprehensive-guide-to-the-grammar-of-graphics-for-effective-visualization-of-multi-dimensional-1f92b4ed4149)*
]

???
- Where `dplyr` provides a grammar for data manipulation, `ggplot2` does the same for plotting
- Up to now, most likely the `graphics` package included in base `R` was your go-to address for crafting visualisations (`plot()`, `hist()`, `boxplot()`).

---

## 4.7 `ggplot2`: Elegant Data Visualisations

```{r, echo=F, out.height='60%', out.width='60%', fig.align='center'}
knitr::include_graphics("https://raw.githubusercontent.com/allisonhorst/stats-illustrations/master/rstats-artwork/ggplot2_masterpiece.png")
```

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Data:** The data set (usually a `tibble`) from which to select the variables that are about to be plotted. It is specified by the first argument in `ggplot()` and thus predestined to be piped into our plot pipeline. 

.pull-left[
**Univariate example:**
```{r p_step_u1, eval=F}
penguins %>% 
  ggplot(data = .)  # equivalent to: ggplot
```
]
.pull-right[
```{r ref.label='p_step_u1', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

.footnote[
*Note: A compact go-to-guide for data visualisations with `ggplot2` is the official [cheat sheet](https://raw.githubusercontent.com/rstudio/cheatsheets/master/data-visualization.pdf).*
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Aesthetics:** Mappings that describe how variables in the data are mapped to aesthetic attributes in the plot, such as axes, shapes, sizes or colors.

.pull-left[
**Univariate example:**
```{r p_step_u2, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm)) 
```
]
.pull-right[
```{r ref.label='p_step_u2', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- You can already see that `ggplot2` extracts the ranges in your variables.

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Geoms:** Geometric objects that determine your overall plot type, e.g., bar, lines, points or boxplots. They specify the graphical representation of your data.

.pull-left[
**Univariate example:**
```{r p_step_u3, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm)) +
  geom_histogram(na.rm = TRUE) 
```
`ggplot2` comes with decent default settings. Each `geom_*()` has its own options for customizing the geom, e.g., 
- change the number of bins with the `bins` argument,
- change the width of the bins with `binwidth` argument.
]
.pull-right[
```{r ref.label='p_step_u3', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
```
]

???
Note that ggplots are constructed by adding layers with `+` instead of ` %>% `

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Geoms:** Geometric objects that determine your overall plot type, e.g., bar, lines, points or boxplots. They specify the graphical representation of your data.

.pull-left[
**Univariate example:**
```{r p_step_u4, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm)) +
  geom_bar(na.rm = TRUE) 
```
Eventually, you may realize the beauty of the `geom_*()` layers. They do all the required calculations for you!

This is due to the frequently overlooked `stat` argument (which defaults to `stat = "count"` for the `geom_bar()` layer).
]
.pull-right[
```{r ref.label='p_step_u4', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Stats:** Statistical transformations provide a summary of the data. They can be used to transform a given variable without changing the plot type (i.e. geom).

.pull-left[
**Univariate example:**
```{r p_step_u5, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm)) +
  geom_bar(na.rm = TRUE, stat = "density") 
```
Most of the time you will just plot the data as-is (`stat = "identity"`). As soon as you require some form of statistical transformation (e.g., count, density or unique) before plotting, the `stat` argument can handle this for you.
]
.pull-right[
```{r ref.label='p_step_u5', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- you can also do all the transformations beforehand using `group_by` and `summarise`

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Stats:** Statistical transformations provide a summary of the data. They can be used to transform a given variable without changing the plot type (i.e. geom).

.pull-left[
**Univariate example:**
```{r p_step_u6, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm)) +
  geom_density(na.rm = TRUE) 
```
If you have to manually change the default setting of the `stat` argument, it is likely that `ggplot2` has implemented a corresponding `geom_*()` already.
]
.pull-right[
```{r ref.label='p_step_u6', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

.footnote[
.pull-left[
*Note: For a great explanation of the inner workings of the `stat` layer, see this [blog post](https://yjunechoe.github.io/posts/2020-09-26-demystifying-stat-layers-ggplot2/) by June Choe.*
]]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Data:** The data set (usually a `tibble`) from which to select the variables that are about to be plotted. It is specified by the first argument in `ggplot()` and thus predestined to be piped into our plot pipeline.

.pull-left[
**Bivariate example:**
```{r p_step_b1, eval=F}
penguins %>% 
  ggplot
```
]
.pull-right[
```{r ref.label='p_step_b1', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Aesthetics:** Mappings that describe how variables in the data are mapped to aesthetic attributes in the plot, such as axes, shapes, sizes or colors.

.pull-left[
**Bivariate example:**
```{r p_step_b2, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm, 
             y = body_mass_g)) 
```
]
.pull-right[
```{r ref.label='p_step_b2', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Geoms:** Geometric objects that determine your overall plot type, e.g., bar, lines, points or boxplots. They specify the graphical representation of your data.

.pull-left[
**Bivariate example:**
```{r p_step_b3, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(na.rm = TRUE) 
```
]
.pull-right[
```{r ref.label='p_step_b3', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

There are multiple ways of changing the color, shape or size aesthetics. Remember that using the `aes()` argument **maps** variable values to your aesthetic. The behavior differs for discrete vs. continuous variables.

.pull-left[
**Bivariate example:**
```{r p_step_b4, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species),
             na.rm = TRUE) 
```
]
.pull-right[
```{r ref.label='p_step_b4', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

There are multiple ways of changing the color, shape or size aesthetics. Remember that using the `aes()` argument **maps** variable values to your aesthetic. The behavior differs for discrete vs. continuous variables.

.pull-left[
**Bivariate example:**
```{r p_step_b5, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = bill_depth_mm),
             na.rm = TRUE) 
```
]
.pull-right[
```{r ref.label='p_step_b5', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

By specifying the `color` argument outside of the `aes()` argument, we **set** the color without considering the values of any other variable.

.pull-left[
**Bivariate example:**
```{r p_step_b6, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(color = "red",
             na.rm = TRUE) 
```
For truly customized colors you may refer to [HTML color codes](https://www.w3schools.com/colors/colors_picker.asp) (also called *hex codes*, e.g., `#ff0000` for red) instead of specifying colors by their [predefined name](http://sape.inf.usi.ch/quick-reference/ggplot2/colour) in `R`.
]
.pull-right[
```{r ref.label='p_step_b6', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
Remember:
- if something is specified inside of `aes` it is mapped, i.e. the characteristic depends on the data
- if something is specified outside of `aes`, it is assigned rather manually
- hexcodes: codes specifying the level of red (first two), green (second two) and blue (last two digits) color intensity

---

## 4.7 `ggplot2`: Elegant Data Visualisations

We can do the same in order to change the `shape` and `size` of our data points. Either by mapping them to the values of another variable or by setting them manually outside of the `aes()` argument.

.pull-left[
**Bivariate example:**
```{r p_step_b7a, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(shape = species),
             size = 4,
             na.rm = TRUE) 
```
`ggplot2` provides 24 available shapes for customizing your plot (see [shape overview](https://ggplot2.tidyverse.org/reference/scale_shape-6.png)).
]
.pull-right[
```{r ref.label='p_step_b7a', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

We can do the same in order to change the `shape` and `size` of our data points. Either by mapping them to the values of another variable or by setting them manually outside of the `aes()` argument.

.pull-left[
**Bivariate example:**
```{r p_step_b7b, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(shape = species,
                 size = bill_depth_mm),
             na.rm = TRUE) 
```
]
.pull-right[
```{r ref.label='p_step_b7b', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Facets:** Facets split the plot into multiple subplots based on the levels of one or more factor variables.

.pull-left[
**Bivariate example:**
```{r p_step_b8, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(shape = species),
             na.rm = TRUE) +
  facet_wrap(~ year) 
```
]
.pull-right[
```{r ref.label='p_step_b8', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Facets:** Facets split the plot into multiple subplots based on the levels of one or more factor variables.

.pull-left[
**Bivariate example:**
```{r p_step_b9, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(shape = species),
             na.rm = TRUE) +
  facet_wrap(~ year + island) 
```
]
.pull-right[
```{r ref.label='p_step_b9', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- lets go back to one of the previous plots

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Scales:** Scales control the aesthetic mappings by overriding the *default* settings. For example, they allow to refine the presentation of x- and y-axis, labels or color palettes ex post.

.pull-left[
**Bivariate example:**
```{r p_step_b10, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species),
             size = 3, na.rm = TRUE) +
  scale_colour_brewer(palette = "Set3") 
```
The family of `scale_colour_*()` functions enables you to adjust the values of your `color` aesthetic (e.g., `scale_colour_brewer()` selects a palette from the famous [ColorBrewer](https://colorbrewer2.org/#type=sequential&scheme=BuGn&n=3) project). 

]
.pull-right[
```{r ref.label='p_step_b10', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- use scales to change the default mappings of `ggplot2` afterwards

---

## 4.7 `ggplot2`: Elegant Data Visualisations

**Scales:** Scales control the aesthetic mappings by overriding the *default* settings. For example, they allow to refine the presentation of x- and y-axis, labels or color palettes ex post.

.pull-left[
**Bivariate example:**
```{r p_step_b11, eval=F}
penguins %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species),
             size = 3, na.rm = TRUE) +
  scale_y_log10()  
```
Or use the `scale_*_log10()` functions to improve the readability of your plot in the presence of high-variance variables.
]
.pull-right[
```{r ref.label='p_step_b11', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- here it doesn't really change a lot due to the absence of outliers (respectively rahter low variations)

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Boxplots for numeric variables
```{r, results='hide'}
penguins_long <- penguins %>% 
  tidyr::pivot_longer(
    cols = contains("mm"),
    names_to = "var", values_to = "val"
  ) %>% 
  tidyr::drop_na()

penguins_long
```
]
.pull-right[
<br>
- Use `dplyr::pivot_longer()` to bring data frame into *long* format.
- Take care of missing values using `dplyr::drop_na()` to avoid error messages.
]

```{r, echo=F}
print(penguins_long, n = 6)
```


---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Boxplots for numeric variables
```{r p_step_o1, eval=F}
penguins_long %>% 
  ggplot(aes(x = var, y = val)) +
  geom_boxplot(na.rm = TRUE) +
  geom_jitter(alpha = 0.2, width = 0.3) 
```
- Use `geom_jitter()` to induce some random noise to the data points to prevent overlapping (alternative to `geom_point()`).
- Control transparency of the respective plot element via the `alpha` aesthetic. 
]
.pull-right[
```{r ref.label='p_step_o1', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Ordered bar chart
```{r p_step_o2, eval=F}
plot <- penguins %>%
  dplyr::count(species) %>%
  dplyr::mutate(prop = n / sum(n)) %>%
  ggplot()

plot +
  geom_col(aes(x = prop, y = species))
```
- You can easily store an `ggplot` object in a user-defined variable.
- Use `dplyr::count()` as shortcut for `group_by()` and `summarise(n = n())`.
- You can either set `aes()` and `data` in `ggplot()` (*global*) or in `geom_*()` (*local*). In the latter case the data and mappings are only active on the *geom*-level.
]
.pull-right[
```{r ref.label='p_step_o2', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

???
- global vs. local: if you want to use different data sets for each layer.
- see that you can easily add new layers to an preexisting `ggplot` object

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Ordered bar chart
```{r p_step_o3, eval=F}
plot +
  geom_col(
    aes(x = prop,
        y = forcats::fct_reorder(species, prop))) +
  scale_x_continuous(
    labels = scales::label_percent(1.))
```
- Use `fct_reorder()` from the `forcats` package to reorder the levels of `species` by their relative frequency (`prop`).
- Finally, `scales::label_percent(1.)` formats the axis as percentages, rounded to percentage points.
]
.pull-right[
```{r ref.label='p_step_o3', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Adjacent bar chart
```{r p_step_o4, eval=F}
penguins %>% 
  ggplot(aes(x = species)) +
  geom_bar(aes(fill = island),
           position = "dodge")
```
]
.pull-right[
```{r ref.label='p_step_o4', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

.footnote[.pull-left[
*Note: `geom_col()` takes a `x` and `y` argument, whereas `geom_bar()` only takes an `x` argument and computes the `y`-quantity internally (e.g., the frequency using `stat = "count"`).*
]]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Stacked bar chart
```{r p_step_o5, eval=F}
penguins %>% 
  ggplot(aes(x = species)) +
  geom_bar(aes(fill = island),
           position = "stack")
```
]
.pull-right[
```{r ref.label='p_step_o5', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Stacked bar chart
```{r p_step_o6, eval=F}
penguins %>% 
  ggplot(
    aes(x = forcats::fct_lump(species, n = 1))) +
  geom_bar(
    aes(fill = island),
    position = "stack")
```
In this crude example we lump together all factor levels except the `n = 1` level(s) with the highest number of observations using `forcats::fct_lump()`.
]
.pull-right[
```{r ref.label='p_step_o6', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** High-quality density plot
```{r p_step_o7, eval=F}
p <- penguins %>% 
  ggplot(aes(x = body_mass_g)) +
  geom_density(aes(fill = species),
               na.rm = T, alpha = 0.4)

p
```
]
.pull-right[
```{r ref.label='p_step_o7', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** High-quality density plot
```{r p_step_o8, eval=F}
breaks <- seq(from = 3000, to = 6000, by = 500)
scales <- scales::label_comma(accuracy = 0.0001)

p <- p +
  scale_x_continuous(breaks = breaks,
                     limits = c(2000, 7000)) +
  scale_y_continuous(labels = scales)

p
```
]
.pull-right[
```{r ref.label='p_step_o8', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** High-quality density plot
```{r p_step_o9, eval=F}
p <- p +
  labs(
    title =
      "Density Function for Three Penguin Species of Palmer Penguins",
    subtitle =
      "Palmer Archipelago (2007-2009)",
    caption =
      "Data: https://github.com/allisonhorst/palmerpenguins",
    x = "Body mass [grams]",
    y = "Statistical density"
  )

p
```
]
.pull-right[
```{r ref.label='p_step_o9', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** High-quality density plot
```{r p_step_o10, eval=F}
p <- p +
  theme_classic()  # otherwise: theme_minimal()

p
```
The `theme` function allows you to customize all elements of your plot which are not immediately related to your data, e.g., titles, labels, fonts, background, or legends.

`ggplot2` also comes with a set of [predefined themes](https://ggplot2.tidyverse.org/reference/ggtheme.html) (`theme_*()`).
]
.pull-right[
```{r ref.label='p_step_o10', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** High-quality density plot
```{r p_step_o11, eval=F, out.width='70%', fig.retina=3, fig.align='center'}
p <- p +
  theme(
    legend.position = "top",
    plot.title =
      element_text(size = 14, face = "bold"),
    plot.subtitle =
      element_text(size = 12),
    plot.caption =
      element_text(size = 10, face = "italic"),
    axis.text.x =
      element_text(size = 10),
    axis.text.y =
      element_blank(),
    axis.title =
      element_text(size = 10),
  )

p
```
]
.pull-right[
```{r ref.label='p_step_o11', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Violin Plot
```{r p_step_o12, eval=F}
penguins %>% 
  ggplot(aes(x = species, y = body_mass_g)) +
  geom_violin(aes(fill = species), na.rm = T) +
  theme_classic()

ggsave("./img/violin-plot.PNG",
       device = "png", dpi = 300)
```
- `geom_violin()` creates a cross-over version of a box-plot and a density plot, particularly suitable for visualizing continuous variables.
- `ggsave()` writes the most recent plot to disk.
]
.pull-right[
```{r ref.label='p_step_o12', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
```
]

---

## 4.7 `ggplot2`: Elegant Data Visualisations

.pull-left[
**Other examples:** Lines of Best Fit
```{r p_step_b13, eval=F}
penguins %>% 
  tidyr::drop_na() %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species)) +
  geom_smooth(method = "lm", se = T)
```
- Use `geom_smooth()` to fit a smooth line to depict the relationship between `x` and `y`.
- For the `method` argument specify one of:
  - *lm* (linear model),
  - *glm* (generalized linear model),
  - *gam* (generalized additive model),
  - *loess* (local regression).
- Set the `se` argument to `TRUE` to obtain standard error bands (i.e. confidence intervals).
]
.pull-right[
.panelset[
.panel[.panel-name[lm]
```{r, echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
penguins %>% 
  tidyr::drop_na() %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species)) +
  geom_smooth(method = "lm", se = T)
```
]
.panel[.panel-name[glm]
```{r, echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
penguins %>% 
  tidyr::drop_na() %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species)) +
  geom_smooth(method = "glm", se = T)
```
]
.panel[.panel-name[gam]
```{r, echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
penguins %>% 
  tidyr::drop_na() %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species)) +
  geom_smooth(method = "gam", se = T)
```
]
.panel[.panel-name[loess]
```{r, echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center', message=F}
penguins %>% 
  tidyr::drop_na() %>% 
  ggplot(aes(x = flipper_length_mm,
             y = body_mass_g)) +
  geom_point(aes(color = species)) +
  geom_smooth(method = "loess", se = T)
```
]
]
]

???
- lm: y follows the normal distribution
- glm: y follows distribution other than normal (e.g., logistic or poisson), but also includes lm (generalization)
- gam: y and x can be exhibit non-linear relationships
- loess: local regression fits the relationship between y and x locally and allows for substantial non-linearities

---

## 4.7 `ggplot2`: Elegant Data Visualisations

By now, there is a whole ecosystem (aka the [ggverse](https://github.com/erikgahner/awesome-ggplot2)) of amazing packages, all created in the spirit of `ggplot2`, which further extend its capabilities:

```{r, echo=F, out.height='40%', out.width='40%', dpi=100, out.extra='style="float:right; padding:10px"'}
knitr::include_graphics(
  "https://tenor.com/view/shocked-po-kung-fu-panda-gif-4255877.gif"
)
```

- `scales`: Scale Functions for Visualization
- `ggtext`: Improved Text Rendering Support for `ggplot2`
- `ggraph`: An Implementation of Grammar of Graphics for Graphs and Networks
- `ggstatsplot`: `ggplot2` Based Plots with Statistical Details
- `plotly`: Create Interactive Web Graphics via `plotly.js`
- `patchwork`: The Composer of Plots
- `ggforce`: Accelerating `ggplot2`
- etc.

???
- ggforce: extension off ggplot2 functionality to allows for more customizations

---

background-image: url(https://raw.githubusercontent.com/ropensci/plotly/master/man/figures/plotly.png)
background-position: 97.5% 2.5%
background-size: 15%
layout: false

## 4.7 `plotly`: Interactive Web Graphics

```{r, out.height='70%', out.width='100%'}
plotly::ggplotly(p)
```

---

background-image: url(https://raw.githubusercontent.com/thomasp85/patchwork/master/man/figures/logo.png)
background-position: 97.5% 2.5%
background-size: 7.5%
layout: false

## 4.7 `patchwork`: The Composer of Plots

.pull-left[
```{r, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
library(patchwork)
p + p + p
```
]
.pull-right[
```{r, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
library(patchwork)
p + (p / p)
```
]

---

background-image: url(https://raw.githubusercontent.com/thomasp85/ggforce/master/man/figures/logo.png)
background-position: 97.5% 2.5%
background-size: 7.5%
layout: false

## 4.7 `ggforce`: Accelerating `ggplot2`

.pull-left[
```{r ggforce, eval=F}
penguins %>% 
  drop_na %>% 
  ggplot(aes(x = .panel_x,
             y = .panel_y,
             col = sex,
             fill = sex)) +
  ggforce::geom_autopoint(alpha = 0.5) +
  ggforce::geom_autohistogram(alpha = 0.5) +
  ggforce::facet_matrix(
    rows = vars(species, island,
                body_mass_g,
                flipper_length_mm),
    switch = "both", layer.diag = 2) +
  theme_bw() +
  theme(axis.text.x = element_text(angle = 90))
```
]
.pull-right[
```{r ref.label='ggforce', echo=F, fig.width=8, fig.asp=0.618, fig.retina=3, fig.align='center'}
```
]