Data Viz

background-image: url("img/DAW.png")
background-position: left
background-size: 50%
class: middle, center,

## .base-blue[Data Visualization]

<br>

### .purple[Kelly McConville]

#### .purple[ Stat 100 | Week 3 | Fall 2022]

]

---

## Announcements

* Remember that the [standard office hours schedule](https://docs.google.com/spreadsheets/d/1eHOdLQGw3mEEvahOM1cKi9p0eu_2TdI6G8b4vdb9hqE/edit?usp=sharing) is now in full swing.

****************************

## Goals for Today

First Segment:

* Motivate data visualizations.
* Develop **language** to talk about the components of a graphic.
* Practice deconstructing graphics.
* Discuss good graphical practices.

]

Second Segment:

* Learn the general structure of `ggplot2`.
* Learn a few standard graphs for numerical/quantitative data:
    + **Histogram**: one numerical variable
    + **Side-by-side boxplot**: one numerical variable and one categorical variable
    + **Side-by-side violin plot**: one numerical variable and one categorical variable

]

---

##  But First...

* Where do we access p-sets?

* First step should always be saving the shared `Rmd` to your home folder.

* `RMarkdown` Workflow:
    * Modify the Rmd file.
    * If the modifications are code, run the code in the console to debug.
    * Knit.
    * Look over the output document.
    * Repeat.

---

# Why construct a graph?

### To **summarize** the data.

### To look for **trends** between variables and make **comparisons**.

### To tell a compelling **story**.

---

# Challenger

* On January 27th, 1986, engineers from Morton Thiokol recommended NASA delay launch of space shuttle *Challenger* due to cold weather.
    * Believed cold weather impacted the o-rings that held the rockets together.
    * Used 13 charts in their argument.

* After a two hour conference call, the engineer's recommendation was overruled due to lack of persuasive evidence and the launch proceeded.

* The Challenger exploded 73 seconds into launch.

---

# Challenger

* Here's one of those charts.

]

]

---

# Challenger

* Here's another one of those charts.

]

]

---

# Challenger

* Here's a graphic I created from [Edward Tufte](https://en.wikipedia.org/wiki/Edward_Tufte)'s data.

]

---

# Challenger

* This adaptation is a recreation of Edward Tufte's graphic.

* For more information on this example and other examples, check out [Tufte's book](https://www.edwardtufte.com/tufte/books_visex).

]

---

## Now let's learn the .orange[Grammar of Graphics].

#### We will use this grammar to:

Decompose and understand existing graphs.

Create our own graphs with the `R` package `ggplot2`.

---

# Grammar of Graphics

* **data**: Data frame that contains the raw data
    + Variables

* **geom**: Geometric **shape** that the data are mapped to.
    + Point, line, bar, text, ...
    
* **aesthetic**: Visual properties of the **geom** 
    + X (horizontal) position, y (vertical) position, color, fill, shape

* **scale**: Controls how data are mapped to the visual values of the aesthetic.
    + EX: particular colors

* **guide**: Legend/key to help user convert visual display back to the data

For right now, we won't focus on the **names** of particular types of graphs (e.g., scatterplot) but on the **elements** of graphs.

---

### Example 1

* What story is the graph telling?
* What are the variables here?
* What **geom** are the variables map to?
* What are the **aesthetic**s of the **geom**?
* Which variable sets the value of that **aesthetic**?
* What additional context does this graphic provide?

]

]

---

### Example 2

]

]

---

### Example 3

]

]

---

## Best Practices: Context

* Think about the **stories/questions** your visualization answers.

* Determine what **context/background information** your viewer needs.

* For context, at a minimum include
    + Axis labels (with units reported).
    + Legends.
    + Data source.

* Visualizing data involves **editorial choices**.
    + What to highlight.
    + What comparisons to make easy to see.
    + What scales to use.

---

## Context Example

]

]

---

##  Best Practices: Order of perception

]

]

---

## Best Practices: Be careful with color.

* Consider color blindness

]

<img src="stat100_wk03mon_files/figure-html/unnamed-chunk-10-1.png" width="576" style="display: block; margin: auto;" />
]

---

## Best Practices: Color Palettes -- Sequential

* [Dude map](https://qz.com/316906/the-dude-map-how-american-men-refer-to-their-bros/)

* Note: Maps are also a great way to provide context!

]

]

---

## Best Practices: Color Palettes -- Diverging

* [Adam Pearce's 2015 NBA Games](https://roadtolarissa.com/nba-minutes/)

]

]

---

## Best Practices: Color Palettes -- Qualitative

* [information is beautiful's Best in Show](https://www.informationisbeautiful.net/visualizations/best-in-show-whats-the-top-data-dog/)

]

]

---

##   Best Practices: Aspect Ratio

* Aspect ratio affects our perception of the rate of change

<div class="figure" style="text-align: center">
<img src="img/aspect_ratio.png" alt="http://socviz.co/lookatdata.html" width="80%" />
<p class="caption">http://socviz.co/lookatdata.html</p>
</div>

---

## Many Ways To Visually Tell A Story

Washington Post's Approach:

[Periscopic's Approach](https://guns.periscopic.com/?year=2013)

---

## Bad Graphics

* It is much easier to make a bad graph than a good graph.

]

<div class="figure" style="text-align: center">
<img src="img/pizza_pie.png" alt="YouGov" width="50%" />
<p class="caption">YouGov</p>
</div>

]

---

## Misleading Graphics

* Ethical data viz: Be careful that your editorial choices don't make your viewer make incorrect conclusions about the data

]

<div class="figure" style="text-align: center">
<img src="img/FLguns.jpg" alt="Modern Data Science with R" width="70%" />
<p class="caption">Modern Data Science with R</p>
</div>

]

---

## Summary Thoughts on Best Practices

* Good graphics are one's where the findings and insights are **obvious** to the viewer.
    + You have power over the patterns your viewer sees and the conclusions they draw.  Use that power ethically.

* Facilitate the **comparisons** that correspond to the research question.

* Add information and key **context**.

* Data visualizations are **not neutral**.

* It is easier to see the differences and similarities between different types of graphics if we learn the **grammar of graphics**.

* Practicing **decomposing** graphics should make it easier for us to **compose** our own graphics.

---

## Two Minute Stretch

---

## Load Necessary Packages

`ggplot2` is part of this collection of data science packages.

```r
# Load necessary packages
library(tidyverse)
```

---

## Data Setting: [Eco-Totem Broadway Bicycle Count](https://data.cambridgema.gov/Transportation-Planning/Eco-Totem-Broadway-Bicycle-Count/q8v9-mcfg)

]

]

---

## Import the [Data](https://data.cambridgema.gov/Transportation-Planning/Eco-Totem-Broadway-Bicycle-Count/q8v9-mcfg)

```r
bike_counter <- read_csv("https://data.cambridgema.gov/api/views/q8v9-mcfg/rows.csv")

# Inspect the data
glimpse(bike_counter)
```

```
## Rows: 245,218
## Columns: 7
## $ DateTime  <chr> "06/24/2015 12:00:00 AM", "06/24/2015 12:15:00 AM", "06/24/2…
## $ Day       <chr> "Wednesday", "Wednesday", "Wednesday", "Wednesday", "Wednesd…
## $ Date      <chr> "06/24/2015", "06/24/2015", "06/24/2015", "06/24/2015", "06/…
## $ Time      <time> 00:00:00, 00:15:00, 00:30:00, 00:45:00, 01:00:00, 01:15:00,…
## $ Total     <dbl> 4, 3, 4, 2, 2, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, …
## $ Westbound <dbl> 1, 3, 3, 2, 2, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, …
## $ Eastbound <dbl> 3, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
```

What does a row represent here?

---

## Inspect the Data

```r
# Look at first few rows
head(bike_counter)
```

```
## # A tibble: 6 × 7
##   DateTime               Day       Date       Time   Total Westbound Eastbound
##   <chr>                  <chr>     <chr>      <time> <dbl>     <dbl>     <dbl>
## 1 06/24/2015 12:00:00 AM Wednesday 06/24/2015 00:00      4         1         3
## 2 06/24/2015 12:15:00 AM Wednesday 06/24/2015 00:15      3         3         0
## 3 06/24/2015 12:30:00 AM Wednesday 06/24/2015 00:30      4         3         1
## 4 06/24/2015 12:45:00 AM Wednesday 06/24/2015 00:45      2         2         0
## 5 06/24/2015 01:00:00 AM Wednesday 06/24/2015 01:00      2         2         0
## 6 06/24/2015 01:15:00 AM Wednesday 06/24/2015 01:15      0         0         0
```

---

## Inspect the Data

```r
# Determine type
# To access one variable: dataset$variable
class(bike_counter$Day)
```

```
## [1] "character"
```

```r
class(bike_counter$Total)
```

```
## [1] "numeric"
```

```r
class(bike_counter)
```

```
## [1] "spec_tbl_df" "tbl_df"      "tbl"         "data.frame"
```

---

## Data Wrangling

**We haven't learned this topic yet.**

**I only included this code for completeness/transparency.**

```r
# Fix Date column to be stored with the date class
library(lubridate)
bike_counter <- mutate(bike_counter, Date = mdy(Date))

# Filter to only include two days in July 2019
july_2019 <- filter(bike_counter,
                    Date %in% c(mdy("07/04/2019"),
                                mdy("07/11/2019")))
# Add an Occasion column
july_2019 <- mutate(july_2019,
                    Occasion = if_else(Date == mdy("07/04/2019"),
                                                    "Fourth of July",
                                                    "Normal Thursday"))
```

---

# Grammar of Graphics

* **data**: Data frame that contains the raw data
    + Variables

* **scale**: Controls how data are mapped to the visual values of the aesthetic.
    + EX: particular colors

* **guide**: Legend/key to help user convert visual display back to the data

---

# `ggplot2` example code

**Guiding Principle**: We will map variables from the **data** to the **aes**thetic attributes (e.g. location, size, shape, color) of **geom**etric objects (e.g. points, lines, bars).

```r
ggplot(data = ---, mapping = aes(---)) +
  geom_---(---) 
```

* There are other layers, such as `scales_---_---()` and `labs()`, but we will wait on those.

---

# Histograms

* Binned counts of data.

* Great for assessing shape.
]

.right-column[
<img src="stat100_wk03mon_files/figure-html/unnamed-chunk-28-1.png" width="576" style="display: block; margin: auto;" />

]

---

# Data Shapes

---

# Histograms

```r
# Create histogram
ggplot(data = july_2019, 
       mapping = aes(x = Total)) +
  geom_histogram()
```

]

]
---

# Histograms

```r
# Create histogram
ggplot(data = july_2019, 
       mapping = aes(x = Total)) +
  geom_histogram(color = "white",
                 fill = "violetred1",
                 bins = 50)
```

#### For aesthetics:

* **mapping** to a variable goes in `aes()`
* **setting** to a specific value goes in the `geom_---()`

]

]

---

# Boxplots

* **Five number summary**: 
    + Minimum
    + First quartile (Q1)
    + Median
    + Third quartile (Q3)
    + Maximum
* Interquartile range (IQR) `$=$` Q3 `$-$` Q1
* Outliers: **unusual** points 
    + Boxplot defines unusual as being beyond `$1.5*IQR$` from `$Q1$` or `$Q3$`.
* Whiskers: reach out to the furthest point that is NOT an outlier

]

]

---

# Boxplots

```r
ggplot(data = july_2019, 
       mapping = aes(x = Occasion, 
                     y = Total)) +
  geom_boxplot()
```

]

]

---

# Boxplots

```r
ggplot(data = july_2019, 
       mapping = aes(x = Occasion, 
                     y = Total)) +
  geom_boxplot(fill = "springgreen1")
```

]

]

---

# Boxplots

```r
ggplot(data = july_2019, 
       mapping = aes(x = Occasion, 
                     y = Total,
                     fill = Occasion)) +
  geom_boxplot()
```

]

]

---

# Boxplots

```r
ggplot(data = july_2019, 
       mapping = aes(x = Occasion, 
                     y = Total,
                     fill = Occasion)) +
  geom_boxplot() +
  guides(fill = "none")
```

]

]

---

# Violin Plots

```r
ggplot(data = july_2019, 
       mapping = aes(x = Occasion, 
                     y = Total,
                     fill = Occasion)) +
  geom_violin() +
  guides(fill = "none")
```

]

]

---

# Boxplot Versus Violin Plots

]

]

---

# Recap: `ggplot2`

```r
library(tidyverse)
ggplot(data = ---, mapping = aes(---)) +
  geom_---(---) 
```