MKT615: Data Storytelling for Marketers

class: center, middle, inverse, title-slide

.title[
# MKT615: Data Storytelling for Marketers
]
.subtitle[
## Lecture 5.3: Data Visualization
]
.author[
### Davide Proserpio
]
.institute[
### Marshall School of Business
]

---

<style type="text/css">
@media print {
  .has-continuation {
    display: block !important;
  }
}
.large4 { font-size: 400% }
.large2 { font-size: 200% }
.small90 { font-size: 90% }
.small75 { font-size: 75% }
</style>

# Table of contents

1. [Checklist](#check)

2. [ggplot2](#ggplot2)

3. [Aesthetic](#aesthetic)

4. [Geoms](#geoms)

5. [Layers](#layers)

6. [More things](#more)

7. [Exercises](#exe)

8. [Readings](#readings)

---
name: check

# Checklist

We will need **ggplot2**, **gapminder**, **dplyr**, **data.table**, **ggthemes**, **gganimate**, **hrbrthemes**, **scales**, **tidyverse**

---
class: inverse, center, middle
name:ggplot2

# ggplot2

---

# Elements of ggplot2

[Hadley Wickham's](http://hadley.nz/) ggplot2 is one of the most popular R packages. 
- It also happens to be built upon some deep visualization theory: i.e. Leland Wilkinson's [*The Grammar of Graphics*](https://www.amazon.com/Grammar-Graphics-Statistics-Computing/dp/0387245448).

There's a lot to say about ggplot2's implementation of this "grammar of graphics" approach, but the three key elements are:

1. Your plot ("the visualization") is linked to your variables ("the data") through various **aesthetic mappings**.

2. Once the aesthetic mappings are defined, you can represent your data in different ways by choosing different **geoms** (i.e. "geometric objects" like points, lines or bars).

3. You build your plot in **layers**.

--

Let's review each element in turn with some actual plots.

---
class: inverse, center, middle
name: aesthetic

# Aesthetics

---

# 1. Aesthetic mappings

```r
p = ggplot(data = gapminder, mapping = aes(x = gdpPercap, y = lifeExp)) 
p = p + geom_point()
p
```

<img src="07-dataviz_files/figure-html/unnamed-chunk-2-1.svg" style="display: block; margin: auto;" />

---

# 1. Aesthetic mappings (cont.)

```r
p = ggplot(data = gapminder, mapping = aes(x = gdpPercap, y = lifeExp)) 
p = p + geom_point() 
p
```

Focus on the top line, which contains the initialising `ggplot()` function call. This function accepts various arguments, including:
- Where the data come from (i.e. `data = gapminder`).
- What the aesthetic mappings are (i.e. `mapping = aes(x = gdpPercap, y = lifeExp)`).
- The `aes()` function gathers together each of the aesthetic mappings used by a layer and passes them to the layer’s mapping argument.

--

In this case, the aesthetic mappings here are pretty simple: They just define an x-axis (GDP per capita) and a y-axis (life expecancy).
- However, aesthetics are a very powerful tool. To get a sense of the power and flexibility that comes with this approach, consider what happens if we add more aesthetics to the plot call.

---

# 1. Aesthetic mappings (cont.)

E.g., let's map **color** and point **size** to two variables:

```r
p = ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop, color = continent))
p = p + geom_point()
p
```

<img src="07-dataviz_files/figure-html/aesthetics2-1.svg" style="display: block; margin: auto;" />

--

Note that I've dropped the "data =" and "mapping =" part of the ggplot call (`ggplot2` knows the order of the arguments.)

---

# 1. Aesthetic mappings (cont.)

E.g., let's map **color** and point **size** to two variables:

<img src="07-dataviz_files/figure-html/aesthetics22-1.svg" style="display: block; margin: auto;" />

- As you can see, ggplot automatically assigns *levels* to the aesthetics and creates a legend for each of them. 
- (For `x` and `y` aesthetics, ggplot2 does not create a legend, but it creates an axis line with tick marks and a label.)

---

# 1. Aesthetic mappings (cont.)

We can specify aesthetic mappings in the geom layer too. Doing so aesthetics are applied only locally (i.e., to the specific geometry).

```r
p = ggplot(gapminder, aes(x = gdpPercap, y = lifeExp)) ## Applicable to all geoms
p = p + geom_point(aes(size = pop, col = continent)) ## Applicable to this geom only
p
```

<img src="07-dataviz_files/figure-html/aesthetics3-1.svg" style="display: block; margin: auto;" />

---

# 1. Aesthetic mappings (cont.)

Oops. What went wrong here?

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point(aes(size = "big", col="black"))
p
```

```
## Warning: Using size for a discrete variable is not advised.
```

<img src="07-dataviz_files/figure-html/aesthetics4-1.svg" style="display: block; margin: auto;" />

--

**Answer: **Aesthetics must be mapped to variables, not descriptions!

---

# 1. Aesthetic mappings (cont.)

Note that outside of the `aes()` function, you can pass values and set the aesthetic **manually**, e.g.,

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point(size = 0.1, col = "blue")
p
```

<img src="07-dataviz_files/figure-html/aesthetics44-1.svg" style="display: block; margin: auto;" />

---
class: inverse, center, middle
name: geoms

# Geoms

---

# 2. Geoms

Once your variable relationships have been defined by the aesthetic mappings, you can invoke and combine different geoms to generate different visualizations.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p = p + geom_smooth(method = "loess")
p
```

```
## `geom_smooth()` using formula = 'y ~ x'
```

<img src="07-dataviz_files/figure-html/geoms1-1.svg" style="display: block; margin: auto;" />

---

# 2. Geoms (cont.)

Aesthetics can be applied differentially across geoms.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point(aes(size = pop, col = continent))
p = p + geom_smooth(method = "loess")
p
```

```
## `geom_smooth()` using formula = 'y ~ x'
```

<img src="07-dataviz_files/figure-html/geoms2-1.svg" style="display: block; margin: auto;" />

---

# 2. Geoms (cont.)

The previous plot provides a good illustration of the power (or effect) that comes from assigning aesthetic mappings "globally" vs "locally" (i.e., in the individual geom layers.)
- Compare: What happens if you run the below code chunk?

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp, size = pop, col = continent))
p = p + geom_point(alpha = 0.3)
p = p + geom_smooth(method = "loess")
p
```

---

# 2. Geoms (cont.)

You can use the same idea to specify different data for each layer.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point(aes(color = year))
p = p + geom_smooth( data = subset(gapminder, year == 2002), se = F)
p
```

```
## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
```

<img src="07-dataviz_files/figure-html/geoms6-1.svg" style="display: block; margin: auto;" />

---

# 2. Geoms (cont.)

Similarly, note that some geoms only accept a subset of mappings. E.g. `geom_density()` doesn't know what to do with the "y" aesthetic mapping.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p + geom_density()
```

```
## Warning: The following aesthetics were dropped during statistical transformation: y.
## ℹ This can happen when ggplot fails to infer the correct grouping structure in
##   the data.
## ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
##   variable into a factor?
```

```
## Error in `geom_density()`:
## ! Problem while setting up geom.
## ℹ Error occurred in the 1st layer.
## Caused by error in `compute_geom_1()`:
## ! `geom_density()` requires the following missing aesthetics: y.
```

---

# 2. Geoms (cont.)

We can fix that by being more careful about how we build the plot.

```r
p = ggplot(data = gapminder) ## i.e. No "global" aesthetic mappings"
p + geom_density(aes(x = gdpPercap, fill = continent), alpha=0.3)
```

<img src="07-dataviz_files/figure-html/geoms5-1.svg" style="display: block; margin: auto;" />

---
# 2. Geoms (cont.)

- ggplot2 provides over 40 geoms, and extension packages provide even more (see https://exts.ggplot2.tidyverse.org/gallery/ for a sampling).
- The best way to get a comprehensive overview is the ggplot2 cheatsheet, which you can find at http://rstudio.com/resources/cheatsheets. 
- To learn more about any single geom, use help: ?geom_smooth.

---
class: inverse, center, middle
name: build

# Layers

---

# 3. Build your plot in layers

We've already seen how we can chain (or "layer") consecutive plot elements.

But it bears repeating: You can build out some truly impressive complexity and transformation of your visualization through this simple layering process.
- You don't have to transform your original data; ggplot2 takes care of all of that.
- For example (see next slide for figure).

```r
p2 =
  ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp)) +
  geom_point(aes(size = pop, col = continent), alpha = 0.3) +
  scale_color_brewer(name = "Continent", palette = "Set1") + ## Different color scale
  scale_size(name = "Population", labels = scales::comma) + ## Different point (i.e. legend) scale
  scale_x_log10(labels = scales::dollar) + ## Switch to logarithmic scale on x-axis. Use dollar units.
  labs(x = "\n Log (GDP per capita)", y = "Life Expectancy\n") + ## Better axis titles
  theme_minimal() ## Try a minimal (b&w) plot theme
```

---

# 3. Build your plot in layers (cont.)

<img src="07-dataviz_files/figure-html/layers2-1.svg" style="display: block; margin: auto;" />

---
class: inverse, center, middle
name: more

# More things

---

# Facets

One way to add additional variables is with aesthetics. Another way, particularly useful for categorical variables, is to split your plot into **facets**, subplots that each display one subset of the data.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p = p + facet_wrap(~ continent, nrow = 2)
p
```

<img src="07-dataviz_files/figure-html/facet-1.svg" style="display: block; margin: auto;" />

---
# Facets (cont.)

To facet your plot on the combination of two variables, add facet_grid() to your plot call.

```r
p = ggplot(data = subset(gapminder, year %in% c(2002,2007)), aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p = p + facet_grid( year ~ continent)
p
```

<img src="07-dataviz_files/figure-html/facetgrid-1.svg" style="display: block; margin: auto;" />

---

# Facets (cont.)

If you prefer to not facet in the rows or columns dimension, use a . instead of a variable name, e.g. `+ facet_grid(. ~ cyl)` (the dot is optional).

```r
p = ggplot(data = subset(gapminder, year %in% c(2002,2007)), aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p = p + facet_grid( ~ continent)
p
```

<img src="07-dataviz_files/figure-html/facetgrid2-1.svg" style="display: block; margin: auto;" />

---

# Statistical transformations

`stat_summary` is very useful to create plots with summary stats w/o altering the original data

```r
p = ggplot(data = gapminder, aes(x = year))
p = p + stat_summary(aes(y=gdpPercap), fun = "mean", geom = "point")
p
```

<img src="07-dataviz_files/figure-html/summary-1.svg" style="display: block; margin: auto;" />
--

In the exercises, we are going to see a few additional ways to do statistical transformations.

---

# Saving plots

`ggsave` is very useful to save plots in any format. Note that for papers, the best format is pdf.

```r
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p
```

```r
ggsave("figures/test.pdf", p, h = 3, w = 5)
```

--

Note that you can also combine more than one plot in one pdf, but you need the `patchwork` library:

```r
library(patchwork)
p = ggplot(data = gapminder, aes(x = gdpPercap, y = lifeExp))
p = p + geom_point()
p1 = p
p2 = p
ggsave("figures/side-by-side.pdf", p1+p2, h = 3, w = 5)
```

---

# There is much more

We have barely scratched the surface of ggplot2's functionality... let alone talked about the entire ecosystem of packages that has been built around it.
- Here's are two quick additional examples

--

Note that you will need to install and load some additional packages if you want to recreate the next two figures on your own machine. A quick way to do this:

```r
if (!require("pacman")) install.packages("pacman")
```

```
## Loading required package: pacman
```

```r
pacman::p_load(hrbrthemes, gganimate)
```

---

# There is much more (cont.)

Simple extension: Use an external package theme.

```r
# library(hrbrthemes)
p2 + theme_modern_rc() + geom_point(aes(size = pop, col = continent), alpha = 0.2)
```

<img src="07-dataviz_files/figure-html/modern_rc_theme-1.svg" style="display: block; margin: auto;" />

---

# There is much more (cont.)

Elaborate extension: Animation!

```r
# library(gganimate)
ggplot(gapminder, aes(gdpPercap, lifeExp, size = pop, colour = country)) +
  geom_point(alpha = 0.7, show.legend = FALSE) +
  scale_colour_manual(values = country_colors) +
  scale_size(range = c(2, 12)) +
  scale_x_log10() +
  facet_wrap(~continent) +
  # Here comes the gganimate specific bits
  labs(title = 'Year: {frame_time}', x = 'GDP per capita', y = 'life expectancy') +
  transition_time(year) +
  ease_aes('linear')
```

---
name: exe

# Exercises

Let's use the ggplot2 `mpg` dataset. What does it contain?

1. Plot the average highway miles per gallon (a dot) by type of car (class)

2. Rename x-axis label to "Type of Car", and y-axis label to "Hwy Miles/Gallon"

3. Now redo the plot but using geom_boxplot (in this case you don't need to summarize the data) and ordering the plot from low to high average hwy mileage per gallon

4. Create a bar plot of the number of cars by type of car (class)
  
  
  
5. Reorder the plot above by count (highest bar first)

6. Create a scatter plot of highway per miles against (x_axis) engine displacement, color the dots by type of car (class)

7. Do the plot above but create a panel for each type of car (use facets)

---
name: readings

# Readings

If you want to do some reading and practice on your own:
- [Chapter 3](https://r4ds.had.co.nz/data-visualisation.html) of *R for Data Science* by Hadley Wickham and Garett Grolemund.
- [*Data Visualization: A Practical Guide*](https://socviz.co/makeplot.html) by Kieran Healy.
- [Designing ggplots](https://designing-ggplots.netlify.com) by Malcom Barrett.