Categorical Data

class: middle, center, title-slide

# Categorical Data
## STAE04: Data Visualization
### Johan Larsson
### The Department of Statistics, Lund University

---

## Visualizing Categorical Data

visualizing categorical data usually comes down to visualizing **proportions**

<div class="figure" style="text-align: center">
<img src="09-categorical-data_files/figure-html/unnamed-chunk-1-1.png" alt="A 'scatter plot' of categorical data; not very appealing." width="2000" />
<p class="caption">A 'scatter plot' of categorical data; not very appealing.</p>
</div>

ggplot2 has **limited** capabilities for visualizing categorical data

---

## Barcharts

simple and readable

```r
library(productplots) # for happy data set
happy <- drop_na(happy)

ggplot(happy, aes(happy)) +
* geom_bar()
```

<div class="figure" style="text-align: center">
<img src="09-categorical-data_files/figure-html/unnamed-chunk-2-1.png" alt="Happiness ratings from the GSS in the US." width="432" />
<p class="caption">Happiness ratings from the GSS in the US.</p>
</div>

---

## Grouped Barchart

good when counts are what matters and group sizes are important, but hard to judge within
categories

```r
ggplot(happy, aes(degree, fill = happy)) +
* geom_bar(position = "dodge")
```

---

## Stacked Barchart

same use case as grouped barcharts, but easier to compare within categories whilst
harder to compare between

```r
ggplot(happy, aes(degree, fill = happy)) +
* geom_bar() # position = "stack" is the default
```

---

## Proportional Stacked Barchart

easy to compare both between and within categories but lose information on category size

```r
ggplot(happy, aes(degree, fill = happy)) +
* geom_bar(position = "fill") +
  ylab("Proportion")
```

---

## Mosaic Plots

a type of stacked barchart that maps category size to 
the width of bars

need to use another package for this: [productplots](https://CRAN.R-project.org/package=productplots) or [ggmosaic](https://CRAN.R-project.org/package=ggmosaic), or [vcd](https://CRAN.R-project.org/package=vcd)

```r
library(productplots)
*prodplot(happy, ~ happy + degree) +
  aes(fill = happy) +
  theme(legend.position = "none") # remove legend
```

---

## Mappings

choice of mappings with categorical data is important

```r
*ggplot(happy, aes(happy, fill = degree)) +
  geom_bar(position = "fill", col = 1) +
  ylab("Proportion")
```

<div class="figure" style="text-align: center">
<img src="09-categorical-data_files/figure-html/unnamed-chunk-7-1.png" alt="The proportional stacked bar chart with a different mapping." width="720" />
<p class="caption">The proportional stacked bar chart with a different mapping.</p>
</div>

pay attention to **which relationship it is you want to display**

---

## Waffle Plots

suitable when there are large differences between categories

again need to use a new package: [waffle](https://CRAN.R-project.org/package=waffle)

<div class="figure" style="text-align: center">
<img src="09-categorical-data_files/figure-html/unnamed-chunk-8-1.png" alt="A waffle plot of the happiness data. Every square represents 100 people. The code is quite complicated (see the source code)." width="720" />
<p class="caption">A waffle plot of the happiness data. Every square represents 100 people. The code is quite complicated (see the source code).</p>
</div>

---

## Euler and Venn Diagrams

useful when relationships between categorical variables is at interest

**proportional** diagrams best but not always possible for more than two categories

.pull-left[

```r
library(eulerr)

combo <- c(
  "happy" = 100, 
  "graduated" = 40,
  "happy&graduated" = 35
)

# fit and inspect
r <- euler(combo)

# plot
plot(r)
```
]

.pull-right[
<img src="09-categorical-data_files/figure-html/unnamed-chunk-9-1.png" width="345.6" style="display: block; margin: auto;" />
]

.footnote[
Disclaimer: Johan Larsson is the author of [eulerr](https://CRAN.R-project.org/package=eulerr).
]

---

## More on Visualizing Categorical Data

much more to learn on visualizations of categorical data

packages [vcd](https://CRAN.R-project.org/package=vcd) and 
[vcdExtra](https://CRAN.R-project.org/package=vcdExtra) offer lots of functionality

<div class="figure" style="text-align: center">
<img src="09-categorical-data_files/figure-html/unnamed-chunk-10-1.png" alt="A plot from the vcd package." width="338.4" />
<p class="caption">A plot from the vcd package.</p>
</div>