DSBA 5122: Visual Analytics

# DSBA 5122: Visual Analytics
## Class 5: Distributions and Uncertainty
### Ryan Wesslen
### September 23, 2019

---

# Why view Distributions? Cairo Ch. 7 & Wilke Ch. 7 - 9

---

.pull-left[
<img src="../images/slides/05-class/cairo.png" width="300px" style="display: block; margin: auto;" />
]
--
.pull-right[
<img src="../images/slides/05-class/cairo2.png" width="280px" style="display: block; margin: auto;" />
]

---

---

---

---

---

---

---

---

---

```r
ggplot(df, aes(x, y)) + geom_boxplot()
```
]
.pull-right[

```r
ggplot(df, aes(x, y)) + geom_violin()
```
]

---

---

---

---

[`ggridges` examples](https://cran.r-project.org/web/packages/ggridges/vignettes/gallery.html)

---

[`ggridges` examples](https://cran.r-project.org/web/packages/ggridges/vignettes/gallery.html)

---

## Uncertainty: Cairo Ch. 10 & Wilke Ch. 16

xkcd

---

---

---

---

---

## Bootstrapping: Within ggplot2

```r
organdata %>% 
  ggplot(aes(x = country, y = donors)) +
 {{stat_summary(fun.y = mean, geom = "point", size = 3) +}} 
* stat_summary(fun.data = mean_cl_boot, geom = "errorbar", width = 0.5) +
  coord_flip() +
  cowplot::theme_cowplot() +
  labs(x = " ", 
       y = "Organ Donations in 000's", 
       title = "Avg Organ Donations (000s) by Country")
```

![](05-class_files/figure-html/unnamed-chunk-26-1.png)

<https://rstudio.cloud/spaces/22733/project/527500>

---

## Bootstrapping: Outside ggplot2

```r
organdata %>%
  group_by(country) %>%
 {{do(as_tibble(bind_rows(Hmisc::smean.cl.boot(.$donors)))) %>%}} # bootstrapping by country
  ggplot(aes(x = reorder(country, Mean), y = Mean)) +
 {{geom_point(size = 3) + }}   # Plot means as geom_point
  geom_errorbar(aes(ymin = Lower, ymax = Upper), width = 0.5) + # Error as geom_errorbar
  coord_flip() +
  cowplot::theme_cowplot() +
  labs(x = " ", y = "Organ Donations in 000's", title = "Avg Organ Donations (000s) by Country")
```

![](05-class_files/figure-html/unnamed-chunk-28-1.png)

<https://rstudio.cloud/spaces/22733/project/527500>

---

# Bootstrapping with HOPs + `gganimate`

![](../../images/slides/05-class/bootstrap-demo-1.gif)

[`ungeviz` package by Claus Wilke](https://wilkelab.org/ungeviz/articles/sampling-bootstrapping.html)
---

# Unemployment Rate

```r
df %>%
  ggplot(aes(x = date, y = unemployment)) +
  geom_line() +
  coord_cartesian(ylim = c(0, .11), expand = FALSE),
  scale_y_continuous(labels = scales::percent) +
  labs(x = NULL, y = NULL, subtitle = "US unemployment over time")
```

![](../../images/slides/05-class/ggplot-unemploy.png)

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/01-unemp-blog.png" width="800px" style="display: block; margin: auto;" />
]

[Kay and Hullman Multiple Views Blog 1](https://medium.com/multiple-views-visualization-research-explained/uncertainty-visualization-explained-67e7a73f031b)

---

# Types of Uncertainty: Reducible and Irreducible

.center2[
<img src="../../images/slides/05-class/02-unemp-blog.png" width="800px" style="display: block; margin: auto;" />
]
.center2[
<img src="../../images/slides/05-class/03-unemp-blog.png" width="800px" style="display: block; margin: auto;" />
]

[Kay and Hullman Multiple Views Blog 1](https://medium.com/multiple-views-visualization-research-explained/uncertainty-visualization-explained-67e7a73f031b)

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/kay0.png" width="800px" style="display: block; margin: auto;" />
]

Source: [Matthew Kay](https://github.com/mjskay/uncertainty-examples/blob/master/us-unemployment.md)

???

Could use a "predictive bar" for the most likely path (draw) and uncertainty around it... fixes the reader to whatever arbitrary interval (95%) the visualization designer chose to display.

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/kay0b.png" width="800px" style="display: block; margin: auto;" />
]

Source: [Matthew Kay](https://github.com/mjskay/uncertainty-examples/blob/master/us-unemployment.md)

???

By showing multiple intervals, we can distinguish between different intervals of uncertainty...

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/kay0c.png" width="800px" style="display: block; margin: auto;" />
]

Source: [Matthew Kay](https://github.com/mjskay/uncertainty-examples/blob/master/us-unemployment.md)

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/kay1.png" width="800px" style="display: block; margin: auto;" />
]

Source: [Matthew Kay](https://github.com/mjskay/uncertainty-examples/blob/master/us-unemployment.md)

---

# Unemployment Rate

.center2[
<img src="../../images/slides/05-class/kay1.gif" width="800px" style="display: block; margin: auto;" />
]

[Hypothetical Outcome Plots (HOP)](https://medium.com/hci-design-at-uw/hypothetical-outcomes-plots-experiencing-the-uncertain-b9ea60d7c740): [`tidybayes`](https://mjskay.github.io/tidybayes/) & [`gganimate`](https://gganimate.com/)

---

## Why is visualizing uncertainty hard?

- Efficient encodings for uncertainty can be hard to find.

- Make sure people understand encodings (what does the plot mean?).

- Perceptual models of probability (e.g., quantile dot plot, HOP).

- Decisions under uncertainty (e.g., [Gigerenzer et al](https://youtu.be/BcMuYhoL38A) or [Monty Hall problem](https://youtu.be/9vRUxbzJZ9Y)).

- Findings may not apply in all contexts.

- Plus, you still have to actually build it!

[Matthew Kay](https://mjskay.github.io/uncertainty-vis-book/ch-why-hard.html)