Data Visualization: Lesson Design

Stage 0 - Assumptions

Audience
- Graduate students and practicing scientists in numerate disciplines from cosmology to economics
  - Lesson content broadly applicable across disciplines
  - Lesson examples from population demography (Gapminder data)
- Must understand basic data management and manipulation
  - Likely worked with spreadsheets and interactive stats tools (SAS)
  - Familiar with tabular data structure
  - Familiar with filtering and summarizing data in groups
  - Understands basic statistics (mean, stdev, lm) to compare groups
- Should have some limited programming experience in R, including:
  - Content from Lessons 00 - 04 of Data Carpentry’s analyzing ecological data in R
  - How to load a CSV (read.csv())
  - Basic dplyr function (filter(), summarize(), etc.)
Constraints
- One half day 0:00-4:00
  - 3:45 teaching time
  - 0:15 coffee break
- Learners use native installs on their own machines
  - Must setup native local install prior to lesson
  - No time is included in schedule for software setup
- Use RStudio for project management and live coding
Data
- Use the Gapminder data throughout
- Break into multiple files by continent
  - GDP, Population, Lifespan, other variables?
  - To make display of output from examples tidier (use Oceana, which is only two lines)
- Choose various data types or vary analysis to include all of the choices for visualization
- Develop examples to show use of multiple data sets?
Focus on dplyr and ggplot2
- Makes lesson usable to novice programmers by avoiding control flow
- Genuine novices are likely to need some introductory support
Exercises will mostly not be “write this code from scratch”
- Focus on decision-making of data structure and visualization
- Lots of short exercises that can reliably be finished in allotted time
- Use MCQs, fill-in-the-blanks, Parsons Problems, “tweak this code”, etc.
Lesson materials
- Notes for instructors and self-study will be written in Markdown
- Learners could be provided with decision tree and solution code for each exercise

Stage 1 - Desired Results

Goals

Learners will be able to navigate Choosing a Good Chart.
- The appropriate graphic should be chosen to respond directly to the scientific question/hypothesis that is to be addressed with the data
- Key decisions
  1. What is the structure of the data/results to communicate?
    - Relationship
    - Comparison
    - Distribution
    - Composition
  2. How many variables or categories need to be conveyed?
    - Is one of those variables time?
  3. How much data is available for each variable or category?
  4. Are absolute or relative values more appropriate to communicate?
Learners will be able to implement the appropriate graphical device using dplyr and ggplot2.
- Basic graphing layers
  - ggplot()
  - aes()
  - geom_line()
  - geom_bar()
  - geom_smooth()
  - geom_point()
  - geom_boxplot()
  - geom_smooth()
- Facet subplots
  - facet_grid()
  - facet_wrap()
- Dataset and aesthetic adjustments (Publication quality figures)
  - scale_manual()
  - lims()
  - labs()
  - guide_legend()
  - theme()

Summative Assessment

Midpoint: Given a research question and some data, choose and generate the best chart.
Final: Evaluate an analysis report generated from code that includes four or five pairs of research questions and charts:
1. Modify some data that is laid out differently (but is still tabular) than what is required by the graphing process
2. Find and fix a chart that is not the best option to address the research question
3. Modify one chart to be publication quality

Essential Questions

How do I…

Read, analyze, and visualize a tabular data set?
Manipulate, summarize, and analyze data to answer a research question?
Choose the best chart to answer my research question?
Generate the best chart to answer my research question?
Produce a publication quality version of my chart?

Learners Will Be Able To…

Interactively generate analysis code and save it in a file
Read tabular data into a data frame
Manipulate and summarize tabular data using dplyr
Generate publication quality charts using ggplot2
Critique others’ choices of charts and esthetics and express how to improve them
Access help(), vignette tutorials, and online documentation
Interpret common error/warning messages from dplyr and ggplot2

Learners Will Know…

That a clearly defined research question is essential to a successful analysis and graphical representation of the results
Where to find documentation on dplyr and ggplot2
How to find out what other graphical software R offers

Stage 2 - Learning Plan

Exercises Note

For coding problems (i.e., Parsons, Faded, Fix the Code), students will be given an output solution (i.e., data.frame, graphic) corresponding to the desired code solution.
Learning plan code solution

Introductions and Learning Objectives

Teaching: 15 min
- Ensure setup issues resolved
- Assess how much dplyr review will be necessary (15 min allotted below)
Exercises: 0 min
- Accounted for in teaching time (no separate exercise)

Data Management in R

Teaching: 10 min
- Import Gapminder data (read.csv())
- Gapminder data structure (str(), head())
Exercises: 5 min
- “How many rows are in the Gapminder Data?” A. 6
  B. 38
  C. 142
  D. 2007
- “How many columns are in the Gapminder Data?” A. 6
  B. 38
  C. 142
  D. 2007

Data Structures

Teaching: 5 min
- Variables and categories
- Values and replicates
- Absolute values vs. relative values
Exercises: 5 min
- Identify data structure features in Gapminder data
- “What years are represented in the Gapminder data? A. 1952-2007
- “Which variable is more appropriate to group by? Why?” A. country B. gdpPercap

`dplyr` Recap

Teaching: 10 min
- select()
- filter()
- group_by()
- summarize()

Exercises: 5 min

“How many countries are in Africa?” Africa <- filter(data, continent=="Africa") A. 5 B. 38 C. 52 D. 142

“Make a data.frame that stores the count of countries per continent?”

count_countries <- data %>%
group_by(continent) %>%
summarize(countries = n())

> count_countries
Source: local data frame [5 x 2]
        
  continent countries
     (fctr)     (int)
1    Africa        52
2  Americas        25
3      Asia        33
4    Europe        30
5   Oceania         2

Scientific Questions and Hypotheses

Teaching: 10 min
- Define: distribution, relationship, comparison, composition
Exercises: 5 min
- Match Column A structure term to Column B question
- Column A
  - Comparison
  - Distribution
  - Relationship
  - Composition
- Column B
  - “How was the population of Africa distributed across the continent by country in 2007?”
  - “What is the relationship of GDP per capita, life expectancy, and population across the world per country in 2007?”
  - “How was the world population distributed among continents in 2007?”
  - “How has the composition of the world population among continents changed from 1952-2007?”

Introduce ‘Choosing a Good Chart’

Teaching: 10 min
- Chart depends on scientific question and data structure
- “How was the population of Africa distributed across the continent by country in 2007?”
Exercises: 5 min
- Add letter labels to ~7 figures on Choosing a Good Chart
- “Match the ‘Scientific Questions and Hypotheses’ exercise answers with the corresponding graphic on Choosing a Good Chart.”

`ggplot`: Grammar of Graphics

Teaching: 10 min
- Aesthetics mapping
- Layers
- Themes

Exercises: 5 min

Parsons Problem with components and form of ggplot2

geom_histogram() +
labs(x = "Log10( 2007 Population )", y = "Count") 
ggplot(Africa, aes(log10(pop_2007))) +

Line Histogram

Teaching: 10 min
- Single variable distribution with many points
- “How was the population of Africa distributed across the continent by country in 2007?”
- geom_histogram(), geom_density()

Exercises: 10 min

Faded Problem

ggplot(Africa, aes(________, ..count..)) +
  geom_density() ________
  labs(x = "Log10(________)", y = "Count")

Publication Quality Figures

Teaching: 5 min
- labs()
- theme()
- Midpoint assessment
Exercises: 5 min
- Faded Problem Continued
  
  ggplot(________) + ________ + labs(________, y = "Count") + theme_classic(base_size = 24, base_family = ________) + theme(axis.title = element_text(size = 36))

Coffee Break 15 min

Bubble Chart

Teaching: 10 min
- Relationship with three variables
- “What is the relationship of GDP per capita, life expectancy, and population across the world per country in 2007?”
- geom_point(), geom_smooth()
- scale_manual()

Exercises: 10 min

Parsons Problem with blanks

ggplot(data, aes(x = ________, y = ________, size = pop_2007)) +
geom_point() +
theme_classic()
geom_smooth() +
labs(x = ________, y = ________, size = "Population", title = "2007") +

Faded Problem

ggplot(data, aes(________)) +
  geom_smooth(method = ________, color = "black", size = 2) +
  geom_point() +
  scale_size(range = c(________)) +
  scale_x_log10() +
  labs(x = "GDP/capita [adjusted US$]", y = "Life Expectancy [years]", 
       size = "Population [millions]", title = "2007") +
  theme_classic(________) +
  theme(axis.title = ________)

Tidy Data Structure

Teaching: 5 min
- Time series data

Exercises: 5 min

Faded Problem: get data.frame near to ‘Faceted Table’

pop_by_continent <- ________ %>%
  gather(year, population, ________) %>%
  separate(________, c("pop", "year"), sep = "_") %>%
  select(continent, country, ________, ________)

Faded Problem: subset data for 2007
```
pop_2007 <- filter(________)
```

Faceted Table of Histogram Bar Charts

Teaching: 5 min
- Comparison among many items/categories
- “How was the world population distributed among continents in 2007?”
- facet_grid()
- lims()

Exercises: 15 min

Fix the Code

ggplot(pop_2007, aes(year)) +
  geom_histogram(binwidth=0.5) +
  scale_x_log10(limits = c(0.1, 10))
  facet_grid(continent ~ .) +
  labs(x = Population [millions], y = Number of countries, 
       title = "2007") +
  theme_bw(base_size = 42, base_family = "Script") +
  theme(axis.title=element_text(size=36))

Stacked Area Chart

Teaching: 5 min
- Composition changing over time
- “How has the composition of the world population among continents changed from 1952-2007?”
- geom_line(), geom_area()
- guide_legend()
Exercises: 15 min
- Final Assessment
  - Given the ‘bad example’, generate a publication quality figure.
  - Students will have to modify code from geom_line() to geom_area() and add labels and theme layers.

Wrap-Up

Teaching: 10 min
- Overview of choosing a good chart and ggplot2
- Other R graphing tools (i.e., interactive)
Exercises: 0 min

Data Visualization: Lesson Design

Stage 0 - Assumptions

Stage 1 - Desired Results

Goals

Summative Assessment

Essential Questions

Learners Will Be Able To…

Learners Will Know…

Stage 2 - Learning Plan

Exercises Note

Introductions and Learning Objectives

Data Management in R

Data Structures

dplyr Recap

Scientific Questions and Hypotheses

Introduce ‘Choosing a Good Chart’

ggplot: Grammar of Graphics

Line Histogram

Publication Quality Figures

Coffee Break 15 min

Bubble Chart

Tidy Data Structure

Faceted Table of Histogram Bar Charts

Stacked Area Chart

Wrap-Up

`dplyr` Recap

`ggplot`: Grammar of Graphics