Introduction

1. Overview

This Week 05 tutorial builds on Weeks 02-04 and focuses on data visualization using R’s powerful ggplot2 package. This tutorial upgrades the lecture concepts from 05_graphs.rmd into hands-on, code-first workflows:

The grammar of graphics and layered approach to visualization
Histograms for understanding distributions
Scatterplots for exploring relationships
Bar graphs for comparing group means
Line graphs for longitudinal and repeated measures data
Data reshaping (wide ↔︎ long format) for visualization
Professional themes and customization
Best practices for effective data presentation

Reference: This tutorial draws heavily from Field, Miles, and Field [1], Discovering Statistics Using R, Chapter 4 on data visualization, with IEEE-style citations throughout.

Upgrade note: This hands-on tutorial extends the Week 05 lecture slides in 05_graphs.rmd with code-first workflows, beginner-friendly explanations, and theoretical grounding from Field et al. [1].

For Beginners: Quick Start

How to run code:
- RStudio: place cursor in a chunk, press Ctrl+Enter (Cmd+Enter on Mac) to run; use Knit to render the full document.
- VS Code: highlight lines and press Ctrl+Enter, or click the Run button above a chunk.
Installing packages: if you see “there is no package called ‘X’”, run install.packages("X") once, then library(X).
If stuck: restart R (RStudio: Session → Restart R), then re-run chunks from the top (setup first).
Reading errors: focus on the last lines of the message; they usually state the actual problem (e.g., object not found, package missing, misspelled column).

2. Learning Objectives

By the end of this tutorial, you will be able to:

Understand the grammar of graphics and the layered structure of ggplot2
Create and interpret histograms to assess data distributions
Build scatterplots to explore bivariate relationships
Construct bar graphs with error bars to compare group means
Generate line graphs for longitudinal and repeated measures data
Reshape data between wide and long formats for visualization
Apply professional themes and customizations to plots
Follow best practices for effective data presentation

3. Required Packages

# Install if needed
# install.packages(c("tidyverse", "ggplot2", "dplyr", "tidyr", "rio", "reshape", "Hmisc", "kableExtra"))

library(tidyverse)
library(ggplot2)
library(dplyr)
library(tidyr)
library(rio)
library(reshape)  # For melt() - note: modern alternative is tidyr::pivot_longer()
library(Hmisc)    # For confidence interval error bars
library(knitr)
library(kableExtra)

cat("\n=== REPRODUCIBLE SEED INFORMATION ===")

## 
## === REPRODUCIBLE SEED INFORMATION ===

cat("\nGenerator Name: Week05 Data Visualization - ANLY500")

## 
## Generator Name: Week05 Data Visualization - ANLY500

cat("\nMD5 Hash:", gen$get_hash())

## 
## MD5 Hash: 1684addbea1f102090becd04b4f27699

cat("\nAvailable Seeds:", paste(seeds[1:5], collapse = ", "), "...\n\n")

## 
## Available Seeds: -328206224, -138599026, 433056330, 714983507, 289841207 ...

Part 1: The Art of Presenting Data

1.1 Why Visualization Matters

What is data visualization? Field et al. [1, Ch. 4] emphasize that graphs are not merely decorative—they are essential tools for understanding data. Before running any statistical test, you should always plot your data to:

Detect patterns that summary statistics might miss
Identify outliers or unusual observations
Check assumptions (normality, linearity, homoscedasticity)
Communicate findings effectively to diverse audiences

“The human visual system is extremely good at detecting patterns, and a well-designed graph can reveal features of the data that would be difficult to detect from tables of numbers alone.” [1, Ch. 4]

1.2 Principles of Good Graphs

Tufte [2] established foundational principles for data visualization that Field et al. [1, Ch. 4] adopt:

1.2.1 Tufte’s Principles

A good graph should:

Show the data clearly and accurately
Induce the reader to think about the substance rather than methodology or graphic design
Avoid distorting what the data have to say
Present many numbers in a small space efficiently
Make large datasets coherent through effective summarization
Encourage comparison of different pieces of data
Reveal data at several levels of detail, from broad overview to fine structure

Tufte’s Principles of Data Visualization [2]
Principle	Good_Example	Bad_Example
Show the data	Scatterplot showing all individual points	Only showing summary statistics
Focus on substance	Clear axis labels, minimal decoration	3D effects, excessive colors, chartjunk
Avoid distortion	Y-axis starts at zero for bar charts	Truncated Y-axis to exaggerate differences
Maximize data-ink ratio	Remove unnecessary gridlines and borders	Heavy gridlines, decorative backgrounds
Make data coherent	Group related data with color or facets	Too many categories without grouping
Encourage comparison	Side-by-side plots or overlaid lines	Separate plots on different scales
Reveal multiple levels	Interactive plots or small multiples	Single aggregated view only

1.2.2 Common Visualization Mistakes

Field et al. [1, Ch. 4] highlight common pitfalls:

1. Misleading Scales

# Example: Truncated Y-axis exaggerates differences
set.seed(seeds[2])
sales_data <- data.frame(
  Quarter = c("Q1", "Q2", "Q3", "Q4"),
  Sales = c(95, 97, 96, 98)
)

par(mfrow = c(1, 2))

# BAD: Truncated axis
barplot(sales_data$Sales, names.arg = sales_data$Quarter,
        main = "BAD: Truncated Y-axis\n(Exaggerates differences)",
        ylab = "Sales (millions)", ylim = c(94, 99),
        col = "lightcoral", border = "darkred")

# GOOD: Full scale
barplot(sales_data$Sales, names.arg = sales_data$Quarter,
        main = "GOOD: Full Y-axis\n(Shows true scale)",
        ylab = "Sales (millions)", ylim = c(0, 100),
        col = "lightgreen", border = "darkgreen")

par(mfrow = c(1, 1))

Interpretation: The left plot (truncated Y-axis) makes Q4 sales appear dramatically higher than Q1, when in reality the difference is only 3% [1, Ch. 4].

2. 3D Charts and Chartjunk

Field et al. [1, Ch. 4] warn against unnecessary 3D effects, which: - Distort perception of magnitudes - Make precise value reading difficult - Add visual clutter without information

3. Overlays and Pattern Confusion

Avoid overlapping bars with different patterns
Use color sparingly and consistently
Ensure sufficient contrast for accessibility

1.3 Programming Suggestion: Code Stacking

When building complex ggplot2 visualizations, stack your code across multiple lines [1, Ch. 4]. This improves readability and makes troubleshooting easier.

# BAD: Everything on one line (hard to read and debug)
ggplot(data, aes(x, y, color = group)) + geom_point() + geom_line() + xlab("X Label") + ylab("Y Label") + theme_minimal()

# GOOD: Stacked code (easy to read and modify)
ggplot(data, aes(x, y, color = group)) +
  geom_point() +
  geom_line() +
  xlab("X Label") +
  ylab("Y Label") +
  theme_minimal()

Why this matters: When you get an error, R will tell you which line failed. With stacked code, you can comment out layers one at a time to isolate the problem [1, Ch. 4].

Part 2: The Grammar of Graphics and ggplot2

2.1 Understanding the Grammar of Graphics

What is the grammar of graphics? Wilkinson [3] introduced a systematic way to describe the components of a graph. Wickham [4] implemented this in ggplot2, which Field et al. [1, Ch. 4] recommend as the standard for R visualization.

2.1.1 Core Components

Every ggplot2 graph has three essential components:

Data: The dataset you want to visualize
Aesthetics (aes): Mappings from data variables to visual properties (x, y, color, size, shape)
Geometries (geom): The type of plot (points, lines, bars, etc.)

# Basic structure
ggplot(data = <DATA>,               # 1. Data
       aes(x = <X_VAR>,             # 2. Aesthetics
           y = <Y_VAR>,
           color = <COLOR_VAR>)) +
  geom_<TYPE>()                     # 3. Geometry

2.1.2 Layered Approach

ggplot2 builds graphs in layers [1, Ch. 4]:

Base layer: Define data and aesthetics with ggplot()
Geometry layers: Add visual representations with geom_*()
Statistics layers: Add statistical transformations with stat_*()
Scale layers: Customize axes and legends with scale_*()
Coordinate layers: Adjust coordinate systems with coord_*()
Theme layers: Control non-data elements with theme_*()

# Demonstrate layered approach
data(mtcars)

# Start with base
p <- ggplot(mtcars, aes(x = wt, y = mpg))

# Layer 1: Just the base (empty plot)
p

# Layer 2: Add points
p + geom_point()

# Layer 3: Add smooth line
p + geom_point() + geom_smooth(method = "lm", se = FALSE)

# Layer 4: Add labels
p + 
  geom_point() + 
  geom_smooth(method = "lm", se = FALSE) +
  labs(title = "Weight vs. Fuel Efficiency",
       x = "Weight (1000 lbs)",
       y = "Miles per Gallon")

Key insight [1, Ch. 4]: By separating data, aesthetics, and geometries, ggplot2 allows you to quickly try different visualizations of the same data by changing only the geom_*() layer.

2.2 Creating a Reusable Theme

Field et al. [1, Ch. 4] recommend creating a custom theme for consistency across all your plots.

# Create a clean, professional theme
cleanup <- theme(
  panel.grid.major = element_blank(),      # Remove major gridlines
  panel.grid.minor = element_blank(),      # Remove minor gridlines
  panel.background = element_blank(),      # Remove background
  axis.line.x = element_line(color = 'black'),  # Black x-axis
  axis.line.y = element_line(color = 'black'),  # Black y-axis
  legend.key = element_rect(fill = 'white'),    # White legend background
  text = element_text(size = 15)           # Larger text for readability
)

# Test the theme
ggplot(mtcars, aes(x = wt, y = mpg)) +
  geom_point(size = 3, color = "steelblue") +
  labs(title = "Custom Theme Example",
       x = "Weight (1000 lbs)",
       y = "Miles per Gallon") +
  cleanup

Practical tip: Save this cleanup theme in your R script and add it to every plot with + cleanup. This ensures all your visualizations have a consistent, professional appearance [1, Ch. 4].

Part 3: Histograms for Understanding Distributions

3.1 What is a Histogram?

Definition [1, Ch. 4]: A histogram plots: - X-axis: Continuous score bins - Y-axis: Frequency (count) of observations in each bin

Why histograms matter: They help you identify: 1. Shape of the distribution (normal, skewed, bimodal) 2. Spread or variation in scores 3. Outliers or unusual values 4. Gaps in the data

3.1.1 Example Dataset: Jiminy Cricket

Field et al. [1, Ch. 4] use a whimsical example to test the Disney philosophy: “Wishing upon a star makes all your dreams come true.”

Study design: - 250 participants randomly assigned to either: - Wish upon a star for 5 years - Work as hard as they can for 5 years - Outcome: Success measured on a 0–4 scale (0 = complete failure, 4 = complete success) - Measurement: Success assessed at baseline (pre) and after 5 years (post)

# Create example data (simulating the Jiminy Cricket study)
set.seed(seeds[3])

cricket <- data.frame(
  ID = 1:250,
  Strategy = rep(c("Wish upon a star", "Work hard"), each = 125),
  Success_Pre = c(
    rnorm(125, mean = 2.0, sd = 0.8),  # Wish group baseline
    rnorm(125, mean = 2.0, sd = 0.8)   # Work group baseline
  ),
  Success_Post = c(
    rnorm(125, mean = 2.1, sd = 0.9),  # Wish group post (minimal change)
    rnorm(125, mean = 3.2, sd = 0.7)   # Work group post (substantial improvement)
  )
)

# Constrain to 0-4 scale
cricket$Success_Pre <- pmin(pmax(cricket$Success_Pre, 0), 4)
cricket$Success_Post <- pmin(pmax(cricket$Success_Post, 0), 4)

head(cricket) %>%
  kable(caption = "Jiminy Cricket Dataset (First 6 Rows)", digits = 2) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Jiminy Cricket Dataset (First 6 Rows)
ID	Strategy	Success_Pre	Success_Post
1	Wish upon a star	1.01	2.08
2	Wish upon a star	1.86	1.65
3	Wish upon a star	1.74	2.41
4	Wish upon a star	1.79	3.21
5	Wish upon a star	0.59	2.70
6	Wish upon a star	2.14	3.09

3.2 Building Histograms with ggplot2

3.2.1 Basic Histogram

# Step 1: Define the base plot
crickethist <- ggplot(data = cricket, aes(x = Success_Pre))

# Step 2: Add histogram layer
crickethist + 
  geom_histogram()

What happened? ggplot2 automatically chose 30 bins and warned us. Let’s customize this [1, Ch. 4].

3.2.2 Customizing Bin Width

# Better: Specify bin width explicitly
crickethist + 
  geom_histogram(binwidth = 0.5, color = 'purple', fill = 'magenta') +
  xlab("Success Pre-Test Score") +
  ylab("Frequency") +
  cleanup

Interpretation: Most participants started with success scores around 2.0 (the middle of the scale), with roughly symmetric distribution [1, Ch. 4].

3.2.3 Comparing Distributions: Festival Hygiene Example

Field et al. [1, Ch. 4] use a memorable example: festival attendees rating their hygiene on each day (0 = “eau de toilet”, 4 = “eau de toilette”).

# Create festival hygiene data
set.seed(seeds[4])

festival <- data.frame(
  ticknumb = 1:200,
  day1 = rnorm(200, mean = 3.5, sd = 0.5),
  day2 = rnorm(200, mean = 2.8, sd = 0.7),
  day3 = rnorm(200, mean = 1.8, sd = 0.9)
)

# Constrain to 0-4 scale
festival$day1 <- pmin(pmax(festival$day1, 0), 4)
festival$day2 <- pmin(pmax(festival$day2, 0), 4)
festival$day3 <- pmin(pmax(festival$day3, 0), 4)

head(festival) %>%
  kable(caption = "Festival Hygiene Dataset (First 6 Rows)", digits = 2) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Festival Hygiene Dataset (First 6 Rows)
ticknumb	day1	day2	day3
1	2.93	3.22	1.87
2	3.11	3.22	1.07
3	4.00	1.41	2.82
4	2.73	3.59	2.67
5	3.39	1.40	2.45
6	4.00	2.10	0.83

# Histogram of Day 1 hygiene
festivalhist <- ggplot(data = festival, aes(x = day1))

festivalhist + 
  geom_histogram(binwidth = 0.5, color = 'blue', fill = 'lightblue') +
  xlab("Day 1 Hygiene Score") +
  ylab("Frequency") +
  cleanup

Interpretation: On Day 1, most festival-goers rated their hygiene highly (scores around 3.5), suggesting they arrived clean [1, Ch. 4].

3.2.4 Overlaying Density Curves

# Add density curve to see distribution shape more clearly
festivalhist + 
  geom_histogram(aes(y = ..density..), binwidth = 0.5, 
                 color = 'blue', fill = 'lightblue', alpha = 0.7) +
  geom_density(color = 'darkblue', linewidth = 1.5) +
  xlab("Day 1 Hygiene Score") +
  ylab("Density") +
  cleanup

Why density curves? They smooth out the histogram’s jaggedness and make it easier to assess normality [1, Ch. 4].

Part 4: Scatterplots for Exploring Relationships

4.1 What is a Scatterplot?

Definition [1, Ch. 4]: A scatterplot displays: - X-axis: One continuous variable (predictor) - Y-axis: Another continuous variable (outcome) - Points: Each observation plotted at (x, y) coordinates

Why scatterplots matter: They reveal: 1. Direction of relationship (positive, negative, none) 2. Strength of relationship (tight cluster vs. wide scatter) 3. Form of relationship (linear, curved, no pattern) 4. Outliers that might distort correlations or regressions

4.2 Example Dataset: Exam Anxiety

Field et al. [1, Ch. 4] use a dataset of 103 students who rated their exam anxiety, time spent revising, exam performance, and gender.

# Create exam anxiety dataset
set.seed(seeds[5])

exam <- data.frame(
  Code = 1:103,
  Revise = runif(103, min = 5, max = 30),  # Hours of revision
  Exam = rnorm(103, mean = 60, sd = 15),   # Exam score
  Anxiety = rnorm(103, mean = 50, sd = 20), # Anxiety score
  Gender = sample(1:2, 103, replace = TRUE)
)

# Create negative correlation between anxiety and exam score
exam$Exam <- 80 - 0.4 * exam$Anxiety + rnorm(103, mean = 0, sd = 10)
exam$Exam <- pmin(pmax(exam$Exam, 0), 100)  # Constrain to 0-100

# Factor gender
exam$Gender <- factor(exam$Gender, levels = c(1, 2), labels = c("Male", "Female"))

head(exam) %>%
  kable(caption = "Exam Anxiety Dataset (First 6 Rows)", digits = 1) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Exam Anxiety Dataset (First 6 Rows)
Code	Revise	Exam	Anxiety	Gender
1	25.3	62.0	60.7	Female
2	11.2	52.4	54.8	Female
3	20.8	69.6	36.8	Female
4	14.0	74.7	8.1	Female
5	17.3	60.9	45.1	Male
6	29.8	55.6	74.4	Male

4.3 Simple Scatterplot

# Basic scatterplot
scatter <- ggplot(exam, aes(x = Anxiety, y = Exam))

scatter +
  geom_point(size = 3, color = "steelblue", alpha = 0.7) +
  xlab("Anxiety Score") +
  ylab("Exam Score (%)") +
  cleanup

Interpretation: There appears to be a negative relationship—as anxiety increases, exam scores tend to decrease [1, Ch. 4].

4.4 Adding a Regression Line

# Add linear regression line with confidence band
scatter +
  geom_point(size = 3, color = "steelblue", alpha = 0.7) +
  geom_smooth(method = 'lm', formula = y ~ x, 
              color = 'black', fill = 'lightblue') +
  xlab('Anxiety Score') +
  ylab('Exam Score (%)') +
  cleanup

Interpretation: The regression line confirms the negative trend. The shaded band shows the 95% confidence interval—wider at the extremes where we have fewer data points [1, Ch. 4].

4.5 Grouped Scatterplot

Question: Does the anxiety-performance relationship differ by gender?

# Scatterplot with groups
scatter2 <- ggplot(exam, aes(x = Anxiety, y = Exam, 
                             color = Gender, fill = Gender))

scatter2 +
  geom_point(size = 3, alpha = 0.7) +
  geom_smooth(method = "lm", formula = y ~ x, alpha = 0.2) +
  xlab("Anxiety Score") +
  ylab("Exam Score (%)") +
  cleanup +
  scale_fill_manual(name = "Gender of Participant",
                    labels = c("Men", "Women"),
                    values = c("purple", "grey")) +
  scale_color_manual(name = "Gender of Participant",
                     labels = c("Men", "Women"),
                     values = c("purple", "grey10"))

Interpretation: Both genders show negative anxiety-performance relationships, with similar slopes. This suggests the effect of anxiety on exam performance is consistent across genders [1, Ch. 4].

4.6 Scatterplot Matrix with GGally

For exploring multiple relationships simultaneously, Field et al. [1, Ch. 4] recommend scatterplot matrices.

library(GGally)

ggpairs(data = exam[, c("Revise", "Anxiety", "Exam", "Gender")],
        title = "Exam Anxiety: Scatterplot Matrix",
        mapping = aes(color = Gender, alpha = 0.5))

Interpretation: - Diagonal: Distributions of each variable by gender - Lower triangle: Scatterplots showing bivariate relationships - Upper triangle: Correlation coefficients

This reveals that revision time has a weak positive correlation with exam scores, while anxiety has a strong negative correlation [1, Ch. 4].

Part 5: Data Reshaping for Visualization

5.1 Wide vs. Long Format

Wide format: Each participant gets one row; repeated measures are in separate columns [1, Ch. 4].

Long format: Each measurement gets one row; a variable indicates which measure it is [1, Ch. 4].

# Wide format example
wide_data <- data.frame(
  ID = 1:5,
  Baseline = c(10, 12, 11, 13, 10),
  Month_6 = c(15, 17, 14, 18, 16),
  Month_12 = c(20, 22, 19, 23, 21)
)

wide_data %>%
  kable(caption = "Wide Format: One Row per Participant") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Wide Format: One Row per Participant
ID	Baseline	Month_6	Month_12
1	10	15	20
2	12	17	22
3	11	14	19
4	13	18	23
5	10	16	21

# Convert to long format
long_data <- wide_data %>%
  pivot_longer(cols = c(Baseline, Month_6, Month_12),
               names_to = "Time",
               values_to = "Score")

long_data %>%
  head(10) %>%
  kable(caption = "Long Format: One Row per Measurement") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Long Format: One Row per Measurement
ID	Time	Score
1	Baseline	10
1	Month_6	15
1	Month_12	20
2	Baseline	12
2	Month_6	17
2	Month_12	22
3	Baseline	11
3	Month_6	14
3	Month_12	19
4	Baseline	13

Why reshape? ggplot2 requires long format for most visualizations because it maps variables to aesthetics—you need a single column for the outcome and a factor column to indicate groups [1, Ch. 4].

5.2 Reshaping the Jiminy Cricket Data

# Original cricket data is wide
head(cricket, 3)

# Reshape to long format
longcricket <- cricket %>%
  pivot_longer(cols = c(Success_Pre, Success_Post),
               names_to = "Time",
               values_to = "Score") %>%
  mutate(Time = factor(Time, 
                       levels = c("Success_Pre", "Success_Post"),
                       labels = c("Baseline", "5 Years")))

longcricket %>%
  head(10) %>%
  kable(caption = "Long Format Cricket Data", digits = 2) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Long Format Cricket Data
ID	Strategy	Time	Score
1	Wish upon a star	Baseline	1.01
1	Wish upon a star	5 Years	2.08
2	Wish upon a star	Baseline	1.86
2	Wish upon a star	5 Years	1.65
3	Wish upon a star	Baseline	1.74
3	Wish upon a star	5 Years	2.41
4	Wish upon a star	Baseline	1.79
4	Wish upon a star	5 Years	3.21
5	Wish upon a star	Baseline	0.59
5	Wish upon a star	5 Years	2.70

Part 6: Bar Graphs for Comparing Group Means

6.1 What is a Bar Graph?

Definition [1, Ch. 4]: A bar graph displays: - X-axis: Categorical groups - Y-axis: Mean of a continuous outcome - Bars: Height represents the group mean - Error bars: Show precision of the mean (confidence interval, standard error, or standard deviation)

Why bar graphs matter: They allow quick visual comparison of means across groups, with error bars indicating whether differences are likely meaningful [1, Ch. 4].

6.1.1 Types of Error Bars

Field et al. [1, Ch. 4] discuss three common types:

Standard Deviation (SD): Shows spread of individual scores
Standard Error (SE): Shows precision of the mean estimate
Confidence Interval (CI): Shows range likely containing the true population mean

Recommendation: Use 95% confidence intervals for inferential purposes—non-overlapping CIs suggest significant differences [1, Ch. 4].

6.2 Example: Chick Flick Study

Field et al. [1, Ch. 4] describe a study testing whether “chick flicks” exist:

Study design: - 20 men and 20 women randomly assigned to watch either: - Bridget Jones’s Diary (presumed “chick flick”) - Memento (control movie) - Outcome: Physiological arousal (higher = more enjoyment)

# Create chick flick dataset
set.seed(seeds[6])

chickflick <- data.frame(
  gender = rep(1:2, each = 20),
  film = rep(rep(1:2, each = 10), 2),
  arousal = c(
    rnorm(10, mean = 15, sd = 5),  # Men watching Bridget Jones
    rnorm(10, mean = 18, sd = 5),  # Men watching Memento
    rnorm(10, mean = 30, sd = 6),  # Women watching Bridget Jones
    rnorm(10, mean = 17, sd = 5)   # Women watching Memento
  )
)

# Factor variables
chickflick$gender <- factor(chickflick$gender,
                            levels = c(1, 2),
                            labels = c("Male", "Female"))

chickflick$film <- factor(chickflick$film,
                         levels = c(1, 2),
                         labels = c("Bridget Jones", "Memento"))

head(chickflick) %>%
  kable(caption = "Chick Flick Dataset (First 6 Rows)", digits = 1) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Chick Flick Dataset (First 6 Rows)
gender	film	arousal
Male	Bridget Jones	17.0
Male	Bridget Jones	13.9
Male	Bridget Jones	15.4
Male	Bridget Jones	10.2
Male	Bridget Jones	13.2
Male	Bridget Jones	26.8

6.3 Bar Chart: One Independent Variable

# Bar chart showing mean arousal by film
chickbar <- ggplot(chickflick, aes(x = film, y = arousal))

chickbar + 
  stat_summary(fun = mean,
               geom = "bar",
               fill = "White", 
               color = "Black") +
  stat_summary(fun.data = mean_cl_normal, 
               geom = "errorbar", 
               width = 0.2) +
  xlab("Film Watched") +
  ylab("Arousal Level") +
  cleanup

Interpretation: On average, participants showed similar arousal levels for both films when gender is ignored. The error bars overlap substantially, suggesting no significant difference [1, Ch. 4].

6.4 Bar Chart: Two Independent Variables

# Bar chart with gender as second factor
chickbar2 <- ggplot(chickflick, aes(x = film, y = arousal, fill = gender))

chickbar2 +
  stat_summary(fun = mean,
               geom = "bar",
               position = "dodge") +
  stat_summary(fun.data = mean_cl_normal,
               geom = "errorbar", 
               position = position_dodge(width = 0.90),
               width = 0.2) +
  xlab("Film Watched") +
  ylab("Arousal Level") + 
  cleanup +
  scale_fill_manual(name = "Gender of Participant", 
                    labels = c("Men", "Women"),
                    values = c("gray30", "gray70"))

Interpretation: Now the pattern is clear! Women showed much higher arousal watching Bridget Jones than Memento, while men showed similar arousal for both films. This supports the “chick flick” hypothesis [1, Ch. 4].

Key insight: The interaction between gender and film was hidden when we ignored gender. This demonstrates why visualizing data by subgroups is crucial [1, Ch. 4].

Part 7: Line Graphs for Longitudinal Data

7.1 When to Use Line Graphs

Field et al. [1, Ch. 4] recommend line graphs when:

X-axis is continuous or ordered (time, dose, age)
Repeated measures or longitudinal data
Tracking change over time or conditions

Why lines instead of bars? Lines emphasize continuity and trends, making them ideal for time-series or dose-response data [1, Ch. 4].

7.2 Example: Hiccup Cures

Field et al. [1, Ch. 4] describe a study testing four hiccup cures:

Study design: - 15 participants tried four interventions: - Baseline (no treatment) - Tongue pulling for 1 minute - Carotid artery massage - Digital rectal massage (yes, really!) - Outcome: Number of hiccups in the minute after each procedure

# Create hiccups dataset
set.seed(seeds[7])

hiccups <- data.frame(
  Participant = 1:15,
  Baseline = rpois(15, lambda = 15),
  Tongue = rpois(15, lambda = 12),
  Carotid = rpois(15, lambda = 8),
  Other = rpois(15, lambda = 5)
)

head(hiccups) %>%
  kable(caption = "Hiccups Dataset (First 6 Rows)") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Hiccups Dataset (First 6 Rows)
Participant	Baseline	Tongue	Carotid	Other
1	15	9	6	4
2	20	17	8	4
3	17	10	6	4
4	12	17	9	6
5	16	7	10	4
6	12	10	7	5

7.3 Reshaping for Line Graphs

# Reshape to long format
longhiccups <- hiccups %>%
  pivot_longer(cols = c(Baseline, Tongue, Carotid, Other),
               names_to = "Intervention",
               values_to = "Hiccups") %>%
  mutate(Intervention = factor(Intervention,
                               levels = c("Baseline", "Tongue", "Carotid", "Other")))

longhiccups %>%
  head(10) %>%
  kable(caption = "Long Format Hiccups Data") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Long Format Hiccups Data
Participant	Intervention	Hiccups
1	Baseline	15
1	Tongue	9
1	Carotid	6
1	Other	4
2	Baseline	20
2	Tongue	17
2	Carotid	8
2	Other	4
3	Baseline	17
3	Tongue	10

7.4 Line Graph: One Independent Variable

# Line graph of mean hiccups by intervention
hiccupline <- ggplot(longhiccups, aes(x = Intervention, y = Hiccups))

hiccupline +
  stat_summary(fun = mean, geom = "point", size = 4) +
  stat_summary(fun = mean, geom = "line", aes(group = 1), linewidth = 1) +
  stat_summary(fun.data = mean_cl_normal, geom = "errorbar", width = 0.2) +
  xlab("Intervention Type") +
  ylab("Number of Hiccups") + 
  cleanup

Interpretation: All three interventions reduced hiccups compared to baseline, with the “Other” intervention (digital rectal massage) being most effective. The aes(group = 1) tells ggplot2 to connect all points with a single line [1, Ch. 4].

7.5 Example: Text Messaging and Grammar

Field et al. [1, Ch. 4] describe a study testing whether text messaging harms grammar:

Study design: - 50 children randomly assigned to: - Text messaging allowed (n = 25) - Text messaging forbidden (n = 25) - Outcome: Grammar test score (% correct) at baseline and 6 months

# Create texting dataset
set.seed(seeds[8])

texting <- data.frame(
  Group = rep(1:2, each = 25),
  Baseline = c(
    rnorm(25, mean = 65, sd = 10),  # Texting allowed
    rnorm(25, mean = 65, sd = 10)   # No texting
  ),
  Six_months = c(
    rnorm(25, mean = 60, sd = 12),  # Texting allowed (decline)
    rnorm(25, mean = 70, sd = 10)   # No texting (improvement)
  )
)

# Factor group
texting$Group <- factor(texting$Group,
                       levels = c(1, 2),
                       labels = c("Texting Allowed", "No Texting Allowed"))

head(texting) %>%
  kable(caption = "Texting Dataset (First 6 Rows)", digits = 1) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Texting Dataset (First 6 Rows)
Group	Baseline	Six_months
Texting Allowed	67.2	76.0
Texting Allowed	64.8	67.8
Texting Allowed	57.0	63.5
Texting Allowed	89.0	64.7
Texting Allowed	68.4	49.2
Texting Allowed	43.4	45.7

7.6 Reshaping for Two-Factor Line Graphs

# Reshape to long format
longtexting <- texting %>%
  pivot_longer(cols = c(Baseline, Six_months),
               names_to = "Time",
               values_to = "Grammar_Score") %>%
  mutate(Time = factor(Time,
                      levels = c("Baseline", "Six_months"),
                      labels = c("Baseline", "Six Months")))

longtexting %>%
  head(10) %>%
  kable(caption = "Long Format Texting Data", digits = 1) %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Long Format Texting Data
Group	Time	Grammar_Score
Texting Allowed	Baseline	67.2
Texting Allowed	Six Months	76.0
Texting Allowed	Baseline	64.8
Texting Allowed	Six Months	67.8
Texting Allowed	Baseline	57.0
Texting Allowed	Six Months	63.5
Texting Allowed	Baseline	89.0
Texting Allowed	Six Months	64.7
Texting Allowed	Baseline	68.4
Texting Allowed	Six Months	49.2

7.7 Line Graph: Two Independent Variables

# Line graph with two factors
textline <- ggplot(longtexting, aes(x = Time, y = Grammar_Score, color = Group))

textline +
  stat_summary(fun = mean, geom = "point", size = 4) +
  stat_summary(fun = mean, geom = "line", aes(group = Group), linewidth = 1) +
  stat_summary(fun.data = mean_cl_normal, geom = "errorbar", width = 0.2) +
  xlab("Measurement Time") +
  ylab("Mean Grammar Score (%)") +
  cleanup +
  scale_color_manual(name = "Texting Option",
                     labels = c("All the texts", "None of the texts"),
                     values = c("black", "grey50")) +
  scale_x_discrete(labels = c("Baseline", "Six Months"))

Interpretation: This reveals a clear interaction effect [1, Ch. 4]: - Texting allowed group: Grammar scores declined from baseline to 6 months - No texting group: Grammar scores improved over the same period

The diverging lines indicate that the effect of time depends on the texting condition—strong evidence that text messaging harms grammar development [1, Ch. 4].

Part 8: Advanced Customization and Best Practices

8.1 Color Scales and Accessibility

Field et al. [1, Ch. 4] emphasize choosing colors that: 1. Are distinguishable by colorblind individuals 2. Print well in grayscale 3. Have sufficient contrast

# Compare color schemes
library(RColorBrewer)

# Display colorblind-friendly palettes
display.brewer.all(colorblindFriendly = TRUE)

# Example with colorblind-friendly palette
ggplot(chickflick, aes(x = film, y = arousal, fill = gender)) +
  stat_summary(fun = mean, geom = "bar", position = "dodge") +
  stat_summary(fun.data = mean_cl_normal, geom = "errorbar", 
               position = position_dodge(0.9), width = 0.2) +
  scale_fill_brewer(palette = "Set2", name = "Gender") +
  xlab("Film") +
  ylab("Arousal Level") +
  cleanup

8.2 Faceting for Small Multiples

What is faceting? Creating multiple subplots for different groups [1, Ch. 4].

# Faceted histogram by gender
ggplot(exam, aes(x = Exam, fill = Gender)) +
  geom_histogram(binwidth = 5, color = "white", alpha = 0.8) +
  facet_wrap(~ Gender, ncol = 1) +
  xlab("Exam Score (%)") +
  ylab("Frequency") +
  cleanup +
  theme(legend.position = "none")

Why faceting? It avoids overplotting and makes comparisons clearer when you have many groups [1, Ch. 4].

8.3 Saving High-Quality Figures

# Save as PNG (for presentations)
ggsave("figure1.png", width = 8, height = 5, dpi = 300)

# Save as PDF (for publications)
ggsave("figure1.pdf", width = 8, height = 5)

# Save as SVG (for vector editing)
ggsave("figure1.svg", width = 8, height = 5)

Recommendation [1, Ch. 4]: Use 300 dpi for print publications, 150 dpi for web/presentations.

8.4 Annotating Plots

# Add annotations to highlight key findings
ggplot(exam, aes(x = Anxiety, y = Exam)) +
  geom_point(size = 3, alpha = 0.6, color = "steelblue") +
  geom_smooth(method = "lm", se = TRUE, color = "black") +
  annotate("text", x = 80, y = 80, 
           label = "High anxiety\n→ Lower scores",
           size = 5, color = "darkred") +
  annotate("segment", x = 75, xend = 85, y = 75, yend = 55,
           arrow = arrow(length = unit(0.3, "cm")), color = "darkred") +
  xlab("Anxiety Score") +
  ylab("Exam Score (%)") +
  cleanup

Part 9: Summary and Best Practices

9.1 Visualization Workflow Checklist

Field et al. [1, Ch. 4] recommend this workflow:

Recommended Visualization Workflow [1, Ch. 4]
Step	Action	Key_Function
1	Explore data with summary statistics	summary(), str(), head()
2	Check for missing values and outliers	is.na(), boxplot()
3	Choose appropriate plot type for your question	See Table 9.1
4	Create basic plot with ggplot2	ggplot(data, aes(…))
5	Add layers (error bars, regression lines, etc.)	geom_*(), + stat_summary()
6	Customize aesthetics (colors, labels, theme)	scale_*(), + labs(), + theme()
7	Check assumptions visually (normality, linearity)	Q-Q plots, residual plots
8	Save high-quality figure with ggsave()	ggsave(‘file.png’, dpi=300)

9.2 Plot Type Selection Guide

Plot Type Selection Guide [1, Ch. 4]
Data_Structure	Research_Question	Recommended_Plot	ggplot2_Geom
One continuous variable	What is the distribution shape?	Histogram + density curve	geom_histogram() + geom_density()
One continuous variable	Are there outliers?	Boxplot	geom_boxplot()
One categorical variable	What are the frequencies?	Bar chart (counts)	geom_bar()
Two continuous variables	Is there a relationship?	Scatterplot ± regression line	geom_point() + geom_smooth()
Continuous outcome by categorical group	Do group means differ?	Bar chart with error bars	stat_summary(geom=‘bar’)
Continuous outcome by categorical group	Do group means differ?	Boxplot or violin plot	geom_boxplot() / geom_violin()
Repeated measures / time series	How do values change over time?	Line graph with error bars	stat_summary(geom=‘line’)
Multiple continuous variables	What are pairwise relationships?	Scatterplot matrix (GGally::ggpairs)	GGally::ggpairs()

9.3 Common ggplot2 Mistakes and Fixes

# ❌ MISTAKE 1: Forgetting aes() for variable mappings
ggplot(data, x = variable, y = variable2) + geom_point()
# ✅ FIX: Wrap variables in aes()
ggplot(data, aes(x = variable, y = variable2)) + geom_point()

# ❌ MISTAKE 2: Using = instead of == in filters
data %>% filter(group = "A")
# ✅ FIX: Use == for logical comparison
data %>% filter(group == "A")

# ❌ MISTAKE 3: Forgetting group aesthetic for line plots
ggplot(data, aes(x = time, y = score, color = group)) + 
  stat_summary(fun = mean, geom = "line")
# ✅ FIX: Add group aesthetic
ggplot(data, aes(x = time, y = score, color = group)) + 
  stat_summary(fun = mean, geom = "line", aes(group = group))

# ❌ MISTAKE 4: Not reshaping data to long format
# Wide format won't work for most ggplot2 visualizations
# ✅ FIX: Use pivot_longer() or melt()
long_data <- wide_data %>% 
  pivot_longer(cols = c(var1, var2), names_to = "variable", values_to = "value")

9.4 Key Takeaways

Based on Field et al. [1, Ch. 4]:

Always plot your data first before running statistical tests
Choose plot types that match your data structure and research question
Use error bars (preferably 95% CIs) to show precision of estimates
Avoid chartjunk (3D effects, excessive colors, decorative elements)
Reshape data to long format for most ggplot2 visualizations
Stack your code across multiple lines for readability
Create a custom theme for consistency across all plots
Check colorblind accessibility and grayscale printing
Save high-resolution figures (300 dpi for print, 150 dpi for web)
Annotate key findings to guide reader interpretation

Part 10: Practice Exercises

Exercise 1: Histogram and Distribution Assessment

Using the mtcars dataset:

Create a histogram of mpg with appropriate bin width
Add a density curve overlay
Assess whether the distribution is approximately normal
Calculate and report skewness and kurtosis

# Your code here
library(moments)

# Histogram with density
ggplot(mtcars, aes(x = mpg)) +
  geom_histogram(aes(y = ..density..), binwidth = 2, 
                 fill = "lightblue", color = "white") +
  geom_density(color = "darkblue", linewidth = 1.5) +
  xlab("Miles per Gallon") +
  ylab("Density") +
  cleanup

# Distribution statistics
cat("Skewness:", skewness(mtcars$mpg), "\n")
cat("Kurtosis:", kurtosis(mtcars$mpg), "\n")

Exercise 2: Scatterplot with Groups

Using the iris dataset:

Create a scatterplot of Sepal.Length vs. Sepal.Width
Color points by Species
Add separate regression lines for each species
Interpret whether the relationship differs by species

# Your code here
ggplot(iris, aes(x = Sepal.Length, y = Sepal.Width, color = Species)) +
  geom_point(size = 3, alpha = 0.7) +
  geom_smooth(method = "lm", se = TRUE) +
  xlab("Sepal Length (cm)") +
  ylab("Sepal Width (cm)") +
  cleanup

Exercise 3: Bar Chart with Error Bars

Using the chickflick dataset:

Create a bar chart showing mean arousal by film
Add 95% confidence interval error bars
Customize the x-axis labels to be more descriptive
Apply a professional theme

# Your code here
ggplot(chickflick, aes(x = film, y = arousal)) +
  stat_summary(fun = mean, geom = "bar", fill = "steelblue") +
  stat_summary(fun.data = mean_cl_normal, geom = "errorbar", width = 0.2) +
  scale_x_discrete(labels = c("Chick Flick", "Action Film")) +
  xlab("Film Type") +
  ylab("Mean Arousal Level") +
  cleanup

Exercise 4: Line Graph with Interaction

Using the longtexting dataset:

Recreate the line graph showing grammar scores over time by group
Add points to show individual means
Customize colors to be colorblind-friendly
Add a title and caption explaining the finding

# Your code here
ggplot(longtexting, aes(x = Time, y = Grammar_Score, color = Group)) +
  stat_summary(fun = mean, geom = "point", size = 4) +
  stat_summary(fun = mean, geom = "line", aes(group = Group), linewidth = 1.5) +
  stat_summary(fun.data = mean_cl_normal, geom = "errorbar", width = 0.2) +
  scale_color_brewer(palette = "Set1") +
  labs(title = "Text Messaging Harms Grammar Development",
       subtitle = "Interaction between time and texting condition",
       x = "Time Point",
       y = "Grammar Score (%)") +
  cleanup

Glossary

Aesthetic (aes): Mapping from data variables to visual properties (x, y, color, size, shape) [1, Ch. 4].

Bar chart: Graph displaying means of a continuous outcome across categorical groups, typically with error bars [1, Ch. 4].

Binwidth: Width of bins in a histogram; smaller values show more detail but may appear noisy [1, Ch. 4].

Chartjunk: Unnecessary decorative elements that distract from data (3D effects, excessive colors, patterns) [2].

Confidence interval (CI): Range of values likely containing the true population parameter; 95% CIs are standard [1, Ch. 2].

Density curve: Smoothed representation of a distribution’s shape, useful for assessing normality [1, Ch. 4].

Error bar: Visual representation of uncertainty (SD, SE, or CI) displayed on bar charts and line graphs [1, Ch. 4].

Faceting: Creating multiple subplots for different groups using facet_wrap() or facet_grid() [1, Ch. 4].

Geometry (geom): The type of visual representation (points, lines, bars) in a ggplot2 layer [1, Ch. 4].

Grammar of graphics: Systematic framework for describing graph components (data, aesthetics, geometries) [3, 4].

Histogram: Graph showing the frequency distribution of a continuous variable [1, Ch. 4].

Interaction effect: When the effect of one variable depends on the level of another variable; revealed by non-parallel lines [1, Ch. 4].

Layer: Building block of ggplot2 graphs; each layer adds a visual or statistical element [1, Ch. 4].

Line graph: Graph connecting means over time or ordered conditions, emphasizing trends [1, Ch. 4].

Long format: Data structure where each measurement gets its own row; required for most ggplot2 visualizations [1, Ch. 4].

Overplotting: When many points overlap, obscuring the true density; solved with transparency (alpha) or jittering [1, Ch. 4].

Scatterplot: Graph showing the relationship between two continuous variables [1, Ch. 4].

Scatterplot matrix: Grid of scatterplots showing pairwise relationships among multiple variables [1, Ch. 4].

Standard error (SE): Standard deviation of the sampling distribution; measures precision of the mean [1, Ch. 2].

Theme: Non-data elements of a plot (fonts, gridlines, backgrounds); customized with theme() [1, Ch. 4].

Wide format: Data structure where each participant gets one row and repeated measures are in separate columns [1, Ch. 4].

References

[1] A. Field, J. Miles, and Z. Field, Discovering Statistics Using R. London: SAGE Publications, 2012.

[2] E. R. Tufte, The Visual Display of Quantitative Information, 2nd ed. Cheshire, CT: Graphics Press, 2001.

[3] L. Wilkinson, The Grammar of Graphics, 2nd ed. New York: Springer, 2005.

[4] H. Wickham, ggplot2: Elegant Graphics for Data Analysis, 2nd ed. New York: Springer, 2016.

[5] W. S. Cleveland, The Elements of Graphing Data, 2nd ed. Summit, NJ: Hobart Press, 1994.

Session Information

sessionInfo()

## R version 4.5.2 (2025-10-31 ucrt)
## Platform: x86_64-w64-mingw32/x64
## Running under: Windows 11 x64 (build 26200)
## 
## Matrix products: default
##   LAPACK version 3.12.1
## 
## locale:
## [1] LC_COLLATE=English_United States.utf8 
## [2] LC_CTYPE=English_United States.utf8   
## [3] LC_MONETARY=English_United States.utf8
## [4] LC_NUMERIC=C                          
## [5] LC_TIME=English_United States.utf8    
## 
## time zone: America/New_York
## tzcode source: internal
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] RColorBrewer_1.1-3 GGally_2.4.0       kableExtra_1.4.0   knitr_1.50        
##  [5] Hmisc_5.2-4        reshape_0.8.10     rio_1.2.4          lubridate_1.9.4   
##  [9] forcats_1.0.1      stringr_1.6.0      dplyr_1.1.4        purrr_1.2.0       
## [13] readr_2.1.6        tidyr_1.3.1        tibble_3.3.0       ggplot2_4.0.1     
## [17] tidyverse_2.0.0    seedhash_0.1.0    
## 
## loaded via a namespace (and not attached):
##  [1] gtable_0.3.6      xfun_0.54         bslib_0.9.0       htmlwidgets_1.6.4
##  [5] lattice_0.22-7    tzdb_0.5.0        vctrs_0.6.5       tools_4.5.2      
##  [9] generics_0.1.4    cluster_2.1.8.1   pkgconfig_2.0.3   Matrix_1.7-4     
## [13] data.table_1.17.8 checkmate_2.3.3   S7_0.2.1          lifecycle_1.0.4  
## [17] compiler_4.5.2    farver_2.1.2      textshaping_1.0.4 htmltools_0.5.8.1
## [21] sass_0.4.10       yaml_2.3.10       htmlTable_2.4.3   Formula_1.2-5    
## [25] pillar_1.11.1     jquerylib_0.1.4   cachem_1.1.0      rpart_4.1.24     
## [29] nlme_3.1-168      ggstats_0.11.0    tidyselect_1.2.1  digest_0.6.37    
## [33] stringi_1.8.7     splines_4.5.2     labeling_0.4.3    fastmap_1.2.0    
## [37] grid_4.5.2        colorspace_2.1-2  cli_3.6.5         magrittr_2.0.4   
## [41] base64enc_0.1-3   foreign_0.8-90    withr_3.0.2       scales_1.4.0     
## [45] backports_1.5.0   timechange_0.3.0  rmarkdown_2.30    nnet_7.3-20      
## [49] gridExtra_2.3     hms_1.1.4         evaluate_1.0.5    viridisLite_0.4.2
## [53] mgcv_1.9-3        rlang_1.1.6       Rcpp_1.1.0        glue_1.8.0       
## [57] xml2_1.5.1        svglite_2.2.2     rstudioapi_0.17.1 jsonlite_2.0.0   
## [61] R6_2.6.1          plyr_1.8.9        systemfonts_1.3.1

End of Week 05: R for Data Analytics Tutorial

ANLY 500 - Analytics I

Harrisburg University

Week 05: R for Data Analytics (Data Visualization with ggplot2)

Ziyuan Huang

Last Updated: December 11, 2025