Lec08: Time

# Lec08: Time
## Stat41: Data Viz
### Prof Amanda Luby
### Swarthmore College

---

# Today:

(1) Return to Regression Discontinuity (5+5)

(2) The Great Debate Pt 1 (15+5)

(3) Interlude: The negative y-axis (5)

(4) The Great Debate Pt 2 (10+5)

---
class: center

<blockquote class="twitter-tweet"><p lang="und" dir="ltr"><a href="https://t.co/nDIJnSNAn7">pic.twitter.com/nDIJnSNAn7</a></p>&mdash; Alex Selby-Boothroyd (@AlexSelbyB) <a href="https://twitter.com/AlexSelbyB/status/1325782481174466562?ref_src=twsrc%5Etfw">November 9, 2020</a></blockquote> <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>

---
# Announcements

(1) "Office Hours" tomorrow

(2) Final Project Milestones

---

# Regression Discontinuity

Return to jamboards/regression discontinuity with the "time" chapter in mind.

+ What is the point of the graph?

+ How does it fail?

+ What would be a better way?

---
class: center, middle

# Recap
---

# The Great Debate Part 1

### Should y axes start at zero?

---
# Examples on jamboards

+ Pike County COVID Cases
+ Average Global Temperature
+ Fox News new cases
+ Fox News Obamacare enrollment

---
class: inverse, center, middle

# Recap

---
# When is it ok to *not* start at zero?

+ When small changes really matter
--

+ When the scale is distorted 
--

+ When values close to zero are impossible!

---
### When small changes really matter

---
# When scale is distorted

---
# When values close to zero are impossible

---

# We **should** start at zero when

+ Area matters (bar chart, histogram)

+ Distance from zero is important

---
# Not everyone agrees on this!

Proof from an email [thread](images/start-at-zero-thread.pdf) of stats professors.

---

### Interlude: the inverted y axis

+ COVID graph on jamboard = misleading

+ FL Gun Death chart = misleading

![](images/FL-gun-deaths.png)

+ But what if it's done well?

---
class:center

![](images/iraqs-bloody-toll.png)
---
class: inverse, middle, center

# The Great Debate Pt 2

### The double y-axis

---
# Example

![](images/double-y-example.png)
---
# Your jamboards:

+ Spurious Correlations
+ COVID Sewage Study
+ Daily High Temperatures
+ UK Kennel Club

---

# Recap

---
class: inverse, middle, center
### When you choose the start and end y-values, you can force the trends to line up however you want!

---
# When can we do it?

+ When two axes measure the same thing

+ When you want to stress me out

---
# Example in R

Don't make me regret this

```r
library(palmerpenguins)
penguin_counts <- penguins %>% 
  group_by(species) %>% 
  summarize(total = n())
total_penguins <- sum(penguin_counts$total)
ggplot(penguin_counts,
       aes(x = species, y = total, 
           fill = species)) +
  geom_col() +
  scale_y_continuous(
    sec.axis = sec_axis(
      trans = ~ . / total_penguins,
      labels = scales::percent)
  ) +
  guides(fill = FALSE) +
  theme_xaringan()
```

![](Lec08_files/figure-html/penguins-double-y-1.png)
]

---

# Preview of next week:

Interactivity + [Animation](https://preview.redd.it/dw0pdfw5icj51.gif?format=mp4&s=92dced4a3f9f774dbaf7d84a8108a966d12f697f)