Lec03: Map Day!

class: center, middle, inverse, title-slide

# Lec03: Map Day!
## Stat41: Data Viz
### Prof Amanda Luby
### Swarthmore College

---

class: center, middle

# Today:

(1) Historical examples

(2) Map basics

(3) Critiques in your groups

(4) Time for Q's

(5) `ggplot2` examples
---

# John Snow's Cholera map

1854: *Cholera* outbreak killed 127 people in 3 days in a London neighborhood

At the time, people thought that Cholera was an airborne disease (now, we know it's waterborne)

John Snow was a physician who was critical of the airborne theory

Let's think about what hospital data might have looked like:

```
## # A tibble: 3 x 6
##   date         last_name first_name address        age cause_of_death
##   <chr>        <chr>     <chr>      <chr>        <dbl> <chr>         
## 1 Aug 31, 1854 Jones     Thomas     26 Broad St.    37 cholera       
## 2 Aug 31, 1854 Jones     Mary       26 Broad St.    11 cholera       
## 3 Sept 1, 1854 Warwick   Martin     14 Broad St.    23 cholera
```
(more in [MDSR](https://mdsr-book.github.io/mdsr2e/ch-spatial.html#motivation-whats-so-great-about-geospatial-data))

---

![John Snow’s original map of the 1854 Broad Street cholera outbreak. Source: Wikipedia.](images/1280px-Snow-cholera-map-1.jpg)

---
class: center, middle

The genius comes *not* from simply plotting the cases, but from *combining* two sets of data that share common features.

(1) Dataset 1: Cholera cases from hospital

(2) Dataset 2: London water pumps

(3) Common features: Latitude and Longitude

---
class: center

John Snow's map (and water pump) are now "famous" among epidemiologists and statisticians.

![John Snow's water pump (and pub) from a 2019 visit](images/IMG_6784.jpeg)

---
class:center

## W.E.B. DuBois: *Visualizing Black America*

![Book Cover](images/dubois-title-page.jpg)
---
When I think about **Insightful** and **Enlightening** data visualizations, these are what come to mind.

![](images/dubois-georgia-land-owned.jpg)

---
class: center

![](images/warnock-win.png)
---
class: inverse, center, middle

# Basics of Maps

---
# Important things to keep in mind:

(1) Coordinate system and projection onto Euclidean space

(2) How data is represented (eg `aesthetics`)

(3) Level of aggregation (eg state vs district vs county vs individuals)

---
class: center, middle

Geospatial data exists on the globe and is generally described with a *latitude* and *longitude*. Any *projection* from the globe to euclidean space (X-Y plane) is going to cause some distortion.

![](images/cairo-projections.png)

---
class: center, middle

![](images/cairo-projection-suggestions.png)
---
![](images/africa-greenland-distortions.png)
---

# How data is represented

+ Chloropleth
+ Proportional symbol 
+ Dot density

---
# Chloropleth (or Thematic maps)

Color or shading shows statistical data on previously-defined regions (e.g. states, countries), not on regions derived from the data.

* Best used for densities or rates, not totals

* Limit number of color classes; higher data values should be "more intense"

* May mislead: large but low-value areas may stand out more than small but high-value areas

* Mapping uncertainty is difficult

Example to keep in mind: election maps

You'll make chloropleth maps in lab today!

---

# Proportional Symbol

Add regions defined by the data to an existing map using their lat/long values

* Better for totals

* Need to be **proportional** (be careful with circular areas)

* Pay attention to symbol placement (town/city vs center of state)

* Sometimes called a "bubble map"

Example to keep in mind: museums

(example code at the end of these slides)
---
class:center, middle

![](Lec03_files/figure-html/unnamed-chunk-3-1.png)

---

# Dot density maps

Add points representing individuals or groups of a fixed size (from the data) to an existing map using their lat/long values

* Can be very beautiful

* Need to be careful with point size (not so big to cover up other points; not too small that you have a mostly-empty graph)

* Choosing what each point represents can mislead

Example to keep in mind: Evergreen forest dot plot (from the Cairo book)

---

![](images/forest-dot-plot.png)

---
class:center, middle

![](images/xkcd-heatmap.png)

---
class: inverse, center, middle

# Group Time!

---

```r
library(ggmap)
museums <- read_csv("https://raw.githubusercontent.com/mateyneykov/315_code_data/master/data/museums.csv") # Read museum data from github
us <- c(left = -125, bottom = 25.75, right = -67, top = 49) #define lat/long box for US map
map <- get_stamenmap(us, zoom = 4, maptype = "toner-lite") # open-source map based on the lat/long coordinates we just defined
nat_hist_museums <- filter(museums, 
                           Type == "NATURAL HISTORY MUSEUM",
                           Income != 0) # filter museum data to Natural History museums only

ggmap(map) + # Plot the open-source map
  geom_point(aes(x = Longitude, y = Latitude, size = Income), 
             data = nat_hist_museums, alpha = .5, col = "darkblue") + # Layer the points from the museum data
  labs(title = "Natural History Museums by Income") + 
  theme_void() # Since we're using a pre-existing image of a map, we don't want any theme elements
```

![](Lec03_files/figure-html/museum_propsymbol-1.png)