Two Variables

class: middle, center, title-slide

.title[
# Two Variables
]
.subtitle[
## Data Visualization
]
.author[
### Johan Larsson
]
.author[
### Behnaz Pirzamanbein
]
.institute[
### The Department of Statistics, Lund University
]

---

## Associations

.pull-left[
How does `\(x\)` relate to `\(y\)`?

It is straightforward when both variables are continuous.

Scatter plots are the bread and butter of visualizations of
two continuous variables.

```r
library(tidyverse)
ggplot(swiss, 
       aes(Examination, 
           Fertility)) +
* geom_point()
```

]

.pull-right[
<div class="figure" style="text-align: center">
<img src="two-variables_files/figure-html/unnamed-chunk-2-1.png" alt="Fertility and % of draftees receiving highest mark on army examination in French-speaking provinces in Switzerland." width="360" />
<p class="caption">Fertility and % of draftees receiving highest mark on army examination in French-speaking provinces in Switzerland.</p>
</div>
]

---

## Patterns

Understanding patterns is key to being able to visualize data effectively.

---

## Variation

.pull-left-40[
A good visualization should describe the variation in the data.
]

.pull-right-60[
<img src="two-variables_files/figure-html/unnamed-chunk-4-1.png" width="396" style="display: block; margin: auto;" />
]

.pull-left-60[
<img src="two-variables_files/figure-html/unnamed-chunk-5-1.png" width="396" style="display: block; margin: auto;" />
]

.pull-right-40[
Sometimes there is a pattern to the variation.
]

---

## Time

Time has a specific order, which makes the line geom a natural choice.

Always place time on the horizontal (x) axis.

```r
ggplot(economics, aes(date, uempmed)) +
* geom_line() +
  xlab("Time") +
  ylab("Unemployment duration")
```

<div class="figure" style="text-align: center">
<img src="two-variables_files/figure-html/unnamed-chunk-6-1.png" alt="Median unemployment duration in the US." width="648" />
<p class="caption">Median unemployment duration in the US.</p>
</div>

---

## Geoms and Time

It is often useful to combine points and lines when measurements are taken at irregular intervals or infrequently.

```r
filter(ChickWeight, Chick == 5) %>%
  ggplot(aes(Time, weight)) +
* geom_point() +
* geom_line()
```

<div class="figure" style="text-align: center">
<img src="two-variables_files/figure-html/unnamed-chunk-7-1.png" alt="Weight of a chick over time." width="432" />
<p class="caption">Weight of a chick over time.</p>
</div>

---

## Categorical Variables

Categorical variables (factors) can be used as the second dimension.

```r
ggplot(msleep, aes(vore, sleep_total)) +
  geom_boxplot()
```

<div class="figure" style="text-align: center">
<img src="two-variables_files/figure-html/unnamed-chunk-8-1.png" alt="Boxplot of total sleep duration for different mammals." width="504" />
<p class="caption">Boxplot of total sleep duration for different mammals.</p>
</div>

---

## Ordered Factors

Pay attention to whether or not the factor is **ordered**.

.pull-left[
<img src="two-variables_files/figure-html/unnamed-chunk-9-1.png" width="360" style="display: block; margin: auto;" />
]

.pull-right[
<img src="two-variables_files/figure-html/unnamed-chunk-10-1.png" width="360" style="display: block; margin: auto;" />
]

---

## Discrete Data

A problem when one (or both) variables are discrete (or categorical)
is that they might **overlap**.

A simple solution: **opacity** (`alpha`)

.pull-left[

```r
ggplot(mpg, aes(cty, hwy)) +
  geom_point() 
```

<img src="two-variables_files/figure-html/unnamed-chunk-11-1.png" width="360" style="display: block; margin: auto;" />
]

.pull-right[

```r
ggplot(mpg, aes(cty, hwy)) +
* geom_point(alpha = 0.2)
```

<img src="two-variables_files/figure-html/unnamed-chunk-12-1.png" width="360" style="display: block; margin: auto;" />
]

---

## Scales

For some variables, it is more natural to study them on a non-standard scale.

Other times it is simply **convenient** (more compact).

.pull-left[

```r
ggplot(msleep, 
       aes(brainwt, 
           sleep_total)) +
  geom_point() +
* scale_y_continuous()
```

<img src="two-variables_files/figure-html/unnamed-chunk-13-1.png" width="360" style="display: block; margin: auto;" />
]

.pull-right[

```r
ggplot(msleep, 
       aes(brainwt, 
           sleep_total)) +
  geom_point() +
* scale_x_log10()
```

<img src="two-variables_files/figure-html/unnamed-chunk-14-1.png" width="360" style="display: block; margin: auto;" />
]

---

## Pairs Plots

Pairs plot plots the pairwise associations of several variables.

It is complicated in ggplot2 (requires **pivoting** and **facets**).

The [GGally](https://ggobi.github.io/ggally) package <a name=cite-emerson2013></a>([Emerson, Green, Schloerke, Crowley, Cook, Hofmann, and Wickham, 2013](https://www.tandfonline.com/doi/ref/10.1080/10618600.2012.694762))
presents a simpler approach.

---

```r
library(GGally)
ggpairs(mpg, columns = c("displ", "drv", "cty", "hwy"))
```

---

## References

<a name=bib-emerson2013></a>[Emerson, J. W., W. A. Green, B. Schloerke,
et al.](#cite-emerson2013) (2013). "The Generalized Pairs Plot". In:
_Journal of Computational and Graphical Statistics_ 22.1, pp. 79-91.
ISSN: 1061-8600. DOI:
[10.1080/10618600.2012.694762](https://doi.org/10.1080%2F10618600.2012.694762).
URL:
[https://www.tandfonline.com/doi/ref/10.1080/10618600.2012.694762](https://www.tandfonline.com/doi/ref/10.1080/10618600.2012.694762)
(visited on Sep. 19, 2020).