5 Introduction to Data Visualization with ggplot2

5.1 Learning Objectives

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

Understand the Grammar of Graphics philosophy and how ggplot2 implements it
Create scatter plots, line plots, bar charts, histograms, box plots, and violin plots
Map variables to aesthetic properties (color, size, shape, alpha)
Apply and customize themes for publication-ready plots
Add labels, titles, and customize scales
Save plots with ggsave() for different output formats
Choose the appropriate plot type for your data and research question

5.2 Introduction to the Grammar of Graphics

5.2.1 What is ggplot2?

ggplot2 is R’s most powerful and popular data visualization package. Created by Hadley Wickham, it implements Leland Wilkinson’s Grammar of Graphics — a systematic way to describe and build any visualization.

Why “gg”plot2?

gg = Grammar of Graphics
2 = Second iteration (there was a ggplot1, but it’s obsolete)

The package is part of the tidyverse and integrates seamlessly with dplyr, tidyr, and other tidyverse tools.

5.2.2 The Grammar of Graphics Philosophy

Traditional plotting systems (like base R graphics) force you to choose from predefined plot types. The Grammar of Graphics takes a different approach: every visualization is built by combining independent components.

flowchart LR
    A[Data] --> B[Aesthetic Mappings]
    B --> C[Geometric Objects]
    C --> D[Statistical Transformations]
    D --> E[Scales]
    E --> F[Coordinate System]
    F --> G[Facets]
    G --> H[Theme]
    H --> I[Final Plot]

5.2.3 The Three Essential Components

Every ggplot2 plot requires three components:

Data: A data frame or tibble
Aesthetic mappings (aes()): How variables map to visual properties (x, y, color, size, etc.)
Geometric objects (geom_*()): The type of plot (points, lines, bars, etc.)

Basic syntax:

ggplot(data = <DATA>) +
  aes(x = <X_VAR>, y = <Y_VAR>) +
  geom_<TYPE>()

Or more commonly:

ggplot(<DATA>, aes(x = <X_VAR>, y = <Y_VAR>)) +
  geom_<TYPE>()

ggplot2 Uses +, Not %>%

Notice we use + to add layers to a ggplot, NOT the pipe (%>% or |>). This is a common mistake!

# CORRECT
ggplot(data, aes(x = var1, y = var2)) +
  geom_point()

# WRONG - don't use pipe!
ggplot(data, aes(x = var1, y = var2)) %>%
  geom_point()

5.2.4 Setup

Let’s load the necessary packages:

library(tidyverse)  # Includes ggplot2, dplyr, etc.

5.3 Your First ggplot: Scatter Plots with `geom_point()`

Scatter plots display the relationship between two continuous variables. Each observation is represented by a point.

5.3.1 Basic Scatter Plot

Let’s create example data on cattle feed intake and weight gain:

# Create example cattle data
cattle <- tibble(
  animal_id = paste0("C", 1001:1030),
  breed = rep(c("Holstein", "Jersey", "Angus"), each = 10),
  feed_intake_kg = c(
    rnorm(10, mean = 25, sd = 3),  # Holstein
    rnorm(10, mean = 18, sd = 2),  # Jersey
    rnorm(10, mean = 22, sd = 2.5) # Angus
  ),
  weight_gain_kg = c(
    rnorm(10, mean = 1.2, sd = 0.15),  # Holstein
    rnorm(10, mean = 0.8, sd = 0.12),  # Jersey
    rnorm(10, mean = 1.0, sd = 0.13)   # Angus
  )
)

# View first few rows
head(cattle)

# A tibble: 6 × 4
  animal_id breed    feed_intake_kg weight_gain_kg
  <chr>     <chr>             <dbl>          <dbl>
1 C1001     Holstein           24.2          0.989
2 C1002     Holstein           22.3          1.27 
3 C1003     Holstein           21.0          1.27 
4 C1004     Holstein           21.8          1.61 
5 C1005     Holstein           27.5          1.40 
6 C1006     Holstein           21.2          1.07

# Create basic scatter plot
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) +
  geom_point()

Reading this code:

ggplot(cattle, ...) — Use the cattle data
aes(x = feed_intake_kg, y = weight_gain_kg) — Map feed intake to x-axis, weight gain to y-axis
geom_point() — Draw points

5.3.2 Adding Color by Group

We can map the breed variable to color:

ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point()

Notice: - Each breed gets a different color automatically - A legend is created automatically - Colors are chosen from a default palette

5.3.3 Customizing Point Aesthetics

Points have several aesthetic properties we can map to variables OR set manually:

Aesthetic	What it controls	Example variable type
`x`	Horizontal position	Continuous
`y`	Vertical position	Continuous
`color`	Point outline color	Categorical or continuous
`size`	Point size	Continuous
`shape`	Point shape	Categorical
`alpha`	Transparency (0-1)	Continuous

Mapping vs Setting

Mapping (inside aes()): Connects a variable to an aesthetic — varies by data

aes(color = breed)  # Color depends on breed variable

Setting (outside aes()): Sets a fixed value for all points

geom_point(color = "blue", size = 3)  # All points blue and size 3

5.3.3.1 Example: Mapping Size

# Map size to feed intake (redundant with x, but demonstrates the concept)
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg,
                   color = breed, size = feed_intake_kg)) +
  geom_point()

5.3.3.2 Example: Setting Fixed Aesthetics

# Make all points larger and semi-transparent
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 4, alpha = 0.6)

5.3.3.3 Example: Using Shape

# Map shape to breed (in addition to color)
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg,
                   color = breed, shape = breed)) +
  geom_point(size = 3)

5.3.4 When to Use Scatter Plots

✅ Use scatter plots when: - You have two continuous variables - You want to explore relationships/correlations - You want to identify outliers or patterns - You want to show individual data points

❌ Don’t use scatter plots when: - One or both variables are categorical (use bar charts or box plots) - You have too many overlapping points (consider hex bins or contour plots) - You want to show change over time for a single entity (use line plots)

5.4 Line Plots with `geom_line()`

Line plots connect points in order, typically used for time series or sequential data.

5.4.1 Example: Growth Curves

# Create growth data for individual animals over time
growth <- tibble(
  animal_id = rep(c("H001", "H002", "H003"), each = 6),
  week = rep(c(0, 2, 4, 6, 8, 10), times = 3),
  weight_kg = c(
    c(45, 58, 72, 88, 105, 122),  # H001
    c(42, 55, 68, 83, 98, 115),   # H002
    c(48, 60, 75, 91, 108, 126)   # H003
  )
)

growth

# A tibble: 18 × 3
   animal_id  week weight_kg
   <chr>     <dbl>     <dbl>
 1 H001          0        45
 2 H001          2        58
 3 H001          4        72
 4 H001          6        88
 5 H001          8       105
 6 H001         10       122
 7 H002          0        42
 8 H002          2        55
 9 H002          4        68
10 H002          6        83
11 H002          8        98
12 H002         10       115
13 H003          0        48
14 H003          2        60
15 H003          4        75
16 H003          6        91
17 H003          8       108
18 H003         10       126

# Create line plot
ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) +
  geom_line()

5.4.2 Combining Lines and Points

Often, it’s helpful to show both the line and the data points:

ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) +
  geom_line() +
  geom_point()

Layer Order Matters

Layers are drawn in the order you add them. Later layers appear on top of earlier layers.

# Points on top of lines (usually preferred)
geom_line() + geom_point()

# Lines on top of points (can obscure data)
geom_point() + geom_line()

5.4.3 Customizing Lines

# Thicker lines with different line types
ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 3)

5.4.4 When to Use Line Plots

✅ Use line plots when: - Data represents a sequence (especially time) - You want to show trends or changes - Points should be connected in a specific order - You’re tracking individual subjects over time

❌ Don’t use line plots when: - Data points are independent (no meaningful order) - You have many overlapping trajectories (hard to read) - X-axis is not continuous or sequential

5.5 Bar Charts with `geom_col()` and `geom_bar()`

Bar charts display categorical data with rectangular bars.

5.5.1 `geom_col()` vs `geom_bar()`

Function	When to use	Y-axis comes from
`geom_col()`	You’ve already calculated the values	Your data
`geom_bar()`	You want to count occurrences	Automatic counting

5.5.2 Using `geom_col()` with Pre-calculated Values

# Average milk production by breed (already summarized)
milk_production <- tibble(
  breed = c("Holstein", "Jersey", "Guernsey", "Brown Swiss"),
  avg_liters_per_day = c(35, 24, 28, 30)
)

milk_production

# A tibble: 4 × 2
  breed       avg_liters_per_day
  <chr>                    <dbl>
1 Holstein                    35
2 Jersey                      24
3 Guernsey                    28
4 Brown Swiss                 30

# Create bar chart
ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) +
  geom_col()

5.5.3 Using `geom_bar()` to Count

# Create dataset with individual animals
animals <- tibble(
  breed = sample(c("Holstein", "Jersey", "Angus"), size = 100, replace = TRUE)
)

# Count animals per breed automatically
ggplot(animals, aes(x = breed)) +
  geom_bar()

geom_bar() Counts for You

When you use geom_bar(), you only specify x (the categorical variable). The y-axis (count) is calculated automatically using stat = "count".

5.5.4 Adding Color

# Color bars by breed
ggplot(milk_production, aes(x = breed, y = avg_liters_per_day, fill = breed)) +
  geom_col()

color vs fill for Bars

color controls the outline of the bar
fill controls the inside of the bar

Usually, you want fill for bar charts:

aes(fill = breed)  # Fills the bars (typical)
aes(color = breed) # Just colors the outlines (usually not what you want)

5.5.5 Grouped and Stacked Bars

# Create data with two categorical variables
feed_trial <- tibble(
  breed = rep(c("Holstein", "Jersey"), each = 3),
  feed_type = rep(c("A", "B", "C"), times = 2),
  avg_gain_kg = c(1.2, 1.4, 1.1, 0.9, 1.0, 0.8)
)

feed_trial

# A tibble: 6 × 3
  breed    feed_type avg_gain_kg
  <chr>    <chr>           <dbl>
1 Holstein A                 1.2
2 Holstein B                 1.4
3 Holstein C                 1.1
4 Jersey   A                 0.9
5 Jersey   B                 1  
6 Jersey   C                 0.8

# Grouped bars
ggplot(feed_trial, aes(x = breed, y = avg_gain_kg, fill = feed_type)) +
  geom_col(position = "dodge")

# Stacked bars (default)
ggplot(feed_trial, aes(x = breed, y = avg_gain_kg, fill = feed_type)) +
  geom_col()

5.5.6 Horizontal Bars

# Use coord_flip() for horizontal bars
ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) +
  geom_col() +
  coord_flip()

5.5.7 When to Use Bar Charts

✅ Use bar charts when: - Comparing quantities across categories - Showing counts or frequencies - Categories are on the x-axis - Bars should start at zero

❌ Don’t use bar charts when: - Showing distributions (use histograms) - Comparing many categories (can be hard to read) - Changes over time are important (use line plots)

5.6 Histograms with `geom_histogram()`

Histograms show the distribution of a single continuous variable by dividing the data into bins and counting how many observations fall into each bin.

5.6.1 Basic Histogram

# Create birth weight data
birth_weights <- tibble(
  calf_id = 1:200,
  birth_weight_kg = rnorm(200, mean = 40, sd = 5)
)

# Create histogram
ggplot(birth_weights, aes(x = birth_weight_kg)) +
  geom_histogram()

Pick the Right Number of Bins!

The default number of bins (30) may not be ideal for your data. Experiment with:

geom_histogram(bins = 20)      # Specify number of bins
geom_histogram(binwidth = 2)   # Specify width of each bin

Too few bins → lose detail; too many bins → noisy

5.6.2 Customizing Binwidth

# Wider bins (3 kg)
ggplot(birth_weights, aes(x = birth_weight_kg)) +
  geom_histogram(binwidth = 3, fill = "steelblue", color = "white")

5.6.3 Overlapping Histograms by Group

# Create breed data with different distributions
set.seed(123)
breeds_vec <- sample(c("Holstein", "Jersey"), size = 200, replace = TRUE, prob = c(0.5, 0.5))

birth_weights_breed <- tibble(
  breed = breeds_vec,
  birth_weight_kg = ifelse(
    breed == "Holstein",
    rnorm(sum(breed == "Holstein"), mean = 42, sd = 5),
    rnorm(sum(breed == "Jersey"), mean = 37, sd = 4)
  )
)

# Overlapping histograms
ggplot(birth_weights_breed, aes(x = birth_weight_kg, fill = breed)) +
  geom_histogram(binwidth = 2, alpha = 0.6, position = "identity")

Use alpha for Overlapping Histograms

When showing multiple distributions, use alpha (transparency) so you can see where they overlap:

geom_histogram(alpha = 0.5, position = "identity")

position = "identity" places histograms on top of each other (default is stacked).

5.6.4 When to Use Histograms

✅ Use histograms when: - Showing the distribution of a continuous variable - Checking for normality - Identifying skewness or outliers - Understanding the shape of your data

❌ Don’t use histograms when: - You have categorical data (use bar charts) - Comparing multiple groups precisely (use box plots or density plots) - You have very few data points (distributions won’t be clear)

5.7 Box Plots with `geom_boxplot()`

Box plots (box-and-whisker plots) display the distribution of data through five summary statistics: minimum, first quartile (Q1), median, third quartile (Q3), and maximum.

5.7.1 Anatomy of a Box Plot

graph TD
    A[Box Plot Components] --> B[Box: IQR Q1 to Q3]
    A --> C[Line in box: Median]
    A --> D[Whiskers: 1.5 × IQR]
    A --> E[Points beyond whiskers: Outliers]

5.7.2 Basic Box Plot

# Create data for different treatment groups
treatments <- tibble(
  treatment = rep(c("Control", "Drug A", "Drug B", "Drug C"), each = 25),
  weight_gain_kg = c(
    rnorm(25, mean = 50, sd = 8),   # Control
    rnorm(25, mean = 58, sd = 7),   # Drug A
    rnorm(25, mean = 62, sd = 9),   # Drug B
    rnorm(25, mean = 55, sd = 6)    # Drug C
  )
)

# Create box plot
ggplot(treatments, aes(x = treatment, y = weight_gain_kg)) +
  geom_boxplot()

5.7.3 Adding Color

ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) +
  geom_boxplot()

5.7.4 Combining Box Plots with Individual Points

It’s often useful to show the underlying data along with the box plot:

# Add jittered points
ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) +
  geom_boxplot(alpha = 0.7) +
  geom_jitter(width = 0.2, alpha = 0.5)

geom_jitter() Reduces Overplotting

When points overlap, geom_jitter() adds small random horizontal noise to make them visible:

geom_jitter(width = 0.2)  # Jitter horizontally (adjust width as needed)

5.7.5 When to Use Box Plots

✅ Use box plots when: - Comparing distributions across groups - You want to show median, quartiles, and outliers - You have multiple groups to compare - You want a compact summary of distributions

❌ Don’t use box plots when: - Your data has multiple modes (bimodal, multimodal) — box plots hide this - You have very few data points (< ~10 per group) - You want to show exact distributions (use violin plots or histograms)

5.8 Violin Plots with `geom_violin()`

Violin plots show the full distribution of data (like a density plot) rotated and mirrored. They’re useful for showing distributions that box plots might hide.

5.8.1 Basic Violin Plot

ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) +
  geom_violin()

5.8.2 Combining Violin Plots with Box Plots

A powerful combination: violin plot shows distribution, box plot shows summary statistics:

ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) +
  geom_violin(alpha = 0.5) +
  geom_boxplot(width = 0.2, fill = "white")

5.8.3 Adding Points

ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) +
  geom_violin(alpha = 0.5) +
  geom_jitter(width = 0.1, alpha = 0.3, size = 1)

5.8.4 When to Use Violin Plots

✅ Use violin plots when: - You want to show full distribution shape - Your data might be bimodal or have multiple peaks - Comparing distributions across groups - You have enough data points (> ~20 per group)

❌ Don’t use violin plots when: - You have very few data points (distributions won’t be smooth) - Readers aren’t familiar with violin plots (less intuitive than box plots) - You only need summary statistics (use box plots)

5.9 Aesthetic Mappings: A Deeper Dive

We’ve seen several aesthetics already. Let’s consolidate what we know:

5.9.1 Common Aesthetics

Aesthetic	Type	Example	Best used with
`x`	Position	`aes(x = treatment)`	All geoms
`y`	Position	`aes(y = weight)`	All geoms
`color`	Color	`aes(color = breed)`	Points, lines, text
`fill`	Fill	`aes(fill = breed)`	Bars, boxes, violins
`size`	Size	`aes(size = age)`	Points, text
`shape`	Shape	`aes(shape = sex)`	Points
`alpha`	Transparency	`aes(alpha = density)`	All geoms
`linetype`	Line type	`aes(linetype = group)`	Lines

5.9.2 Continuous vs Discrete Variables

ggplot2 handles continuous and discrete variables differently:

# Create data with both variable types
mixed_data <- tibble(
  group = rep(c("A", "B", "C"), each = 20),  # Discrete
  score = rnorm(60, mean = 100, sd = 15),    # Continuous
  intensity = rep(1:20, times = 3)           # Continuous
)

# Color by discrete variable (gets distinct colors)
ggplot(mixed_data, aes(x = intensity, y = score, color = group)) +
  geom_point(size = 3)

# Color by continuous variable (gets gradient)
ggplot(mixed_data, aes(x = intensity, y = score, color = intensity)) +
  geom_point(size = 3)

Choosing the Right Aesthetic

Categorical data: Use color, fill, shape, linetype
Continuous data: Use x, y, size, alpha, color (gradient)
Too many categories? (> ~7): Don’t use shape (hard to distinguish); consider faceting instead

5.10 Working with Scales

Scales control how data values map to visual properties. Every aesthetic has a scale.

5.10.1 Scale Functions

Scale functions follow the pattern: scale_<aesthetic>_<type>()

Examples: - scale_x_continuous() — Continuous x-axis - scale_y_log10() — Log10 transform y-axis - scale_color_manual() — Manually specify colors - scale_fill_brewer() — Use ColorBrewer palettes

5.10.2 Modifying Continuous Scales

# Create sample data
growth_rate <- tibble(
  animal_id = 1:30,
  weight_kg = runif(30, min = 400, max = 700),
  gain_per_day_kg = runif(30, min = 0.8, max = 1.5)
)

# Modify axis limits and breaks
ggplot(growth_rate, aes(x = weight_kg, y = gain_per_day_kg)) +
  geom_point() +
  scale_x_continuous(limits = c(350, 750), breaks = seq(350, 750, by = 100)) +
  scale_y_continuous(limits = c(0, 2), breaks = seq(0, 2, by = 0.5))

5.10.3 Log Transformations

# Data with exponential relationship
bacteria <- tibble(
  time_hours = 0:24,
  count = 100 * 2^(0:24 / 4)  # Exponential growth
)

# Regular scale (hard to see early values)
ggplot(bacteria, aes(x = time_hours, y = count)) +
  geom_line() +
  geom_point()

# Log scale (easier to see pattern)
ggplot(bacteria, aes(x = time_hours, y = count)) +
  geom_line() +
  geom_point() +
  scale_y_log10()

5.10.4 Discrete Scales: Manual Colors

# Specify custom colors
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  scale_color_manual(values = c("Holstein" = "blue",
                                  "Jersey" = "orange",
                                  "Angus" = "darkred"))

5.10.5 Using ColorBrewer Palettes

# Use ColorBrewer palette (colorblind-friendly)
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  scale_color_brewer(palette = "Set1")

Colorblind-Friendly Palettes

Use scale_color_viridis_d() (discrete) or scale_color_viridis_c() (continuous) for accessible colors:

scale_color_viridis_d()  # Discrete variable
scale_color_viridis_c()  # Continuous variable

5.11 Labels and Titles with `labs()`

Use labs() to add or modify plot labels:

ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  labs(
    title = "Relationship Between Feed Intake and Weight Gain",
    subtitle = "Data from 30 cattle across three breeds",
    x = "Daily Feed Intake (kg)",
    y = "Daily Weight Gain (kg)",
    color = "Breed",
    caption = "Source: Fictional farm data (2024)"
  )

Label Arguments in labs()

title — Main plot title
subtitle — Secondary title (smaller font)
x, y — Axis labels
color, fill, shape, size — Legend titles (match the aesthetic name)
caption — Small text at bottom (often for data source or notes)

5.12 Themes: Customizing Plot Appearance

Themes control the non-data elements of your plot (background, grid lines, fonts, etc.).

5.12.1 Built-in Themes

ggplot2 includes several complete themes:

# Create base plot
p <- ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  labs(title = "Feed Intake vs Weight Gain")

# Default theme (theme_gray)
p

# theme_minimal() - Clean and modern
p + theme_minimal()

# theme_classic() - No grid lines
p + theme_classic()

# theme_bw() - Black and white
p + theme_bw()

# theme_light() - Light gray grid
p + theme_light()

5.12.2 Setting a Default Theme

You can set a theme for all plots in your session:

# Set default theme for all subsequent plots
theme_set(theme_minimal())

5.12.3 Customizing Themes

Use theme() to modify specific elements:

ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  labs(title = "Feed Intake vs Weight Gain") +
  theme_minimal() +
  theme(
    plot.title = element_text(size = 16, face = "bold"),
    axis.title = element_text(size = 12),
    legend.position = "bottom"
  )

Common Theme Customizations

theme(
  plot.title = element_text(size = 16, face = "bold"),
  axis.text = element_text(size = 10),
  legend.position = "bottom",  # or "top", "left", "right", "none"
  panel.grid.major = element_line(color = "gray90"),
  panel.grid.minor = element_blank()
)

5.13 Coordinate Systems

Coordinate systems control how the x and y axes are rendered.

5.13.1 `coord_cartesian()`: Zooming

coord_cartesian() is useful for zooming without removing data:

# Zoom in on a specific region
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  coord_cartesian(xlim = c(18, 26), ylim = c(0.8, 1.2))

coord_cartesian() vs scale_*_continuous(limits = ...)

coord_cartesian(xlim = c(a, b)) — Zooms without removing data (stat calculations use all data)
scale_x_continuous(limits = c(a, b)) — Removes data outside limits before plotting

Usually, you want coord_cartesian() for zooming.

5.13.2 `coord_flip()`: Horizontal Plots

We saw this earlier with bar charts. It swaps x and y:

ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) +
  geom_col(fill = "steelblue") +
  coord_flip()

5.13.3 Fixed Aspect Ratio

coord_fixed() ensures equal scales on both axes (1 unit on x = 1 unit on y visually):

# Without fixed aspect ratio
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) +
  geom_point()

# With fixed aspect ratio
ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) +
  geom_point() +
  coord_fixed(ratio = 20)  # 20 kg of feed = 1 kg weight gain visually

5.14 Saving Plots with `ggsave()`

After creating a plot, save it with ggsave():

# Create plot
my_plot <- ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) +
  geom_point(size = 3) +
  labs(title = "Feed Intake vs Weight Gain") +
  theme_minimal()

# Save to file
ggsave("cattle_plot.png", plot = my_plot, width = 8, height = 6, dpi = 300)

5.14.1 Common Arguments

Argument	What it does	Example
`filename`	File path and name	`"plots/figure1.png"`
`plot`	ggplot object (defaults to last plot)	`my_plot`
`width`	Width in inches (or cm, mm)	`8`
`height`	Height in inches (or cm, mm)	`6`
`dpi`	Resolution (dots per inch)	`300` (publication quality)
`units`	Units for width/height	`"in"`, `"cm"`, `"mm"`

5.14.2 File Formats

ggplot2 supports many formats (determined by file extension):

PNG (.png) — Good for presentations, web (raster)
PDF (.pdf) — Best for publications (vector, scalable)
JPEG (.jpg) — Compressed (raster, smaller files)
SVG (.svg) — Vector format for web
TIFF (.tiff) — High quality (raster, large files)

# Save as PDF (recommended for publications)
ggsave("figure1.pdf", width = 7, height = 5)

# Save as high-resolution PNG
ggsave("figure1.png", width = 7, height = 5, dpi = 300)

Publication Guidelines

Most journals require: - 300 dpi minimum resolution - PDF or TIFF format for final submission - Specific dimensions (check journal guidelines) - Readable fonts (usually at least 8-10 pt in final figure)

Start with:

ggsave("figure.pdf", width = 7, height = 5, dpi = 300)

5.15 Summary: Choosing the Right Plot Type

Data Type	Goal	Recommended Plot
2 continuous variables	Show relationship	Scatter plot (`geom_point`)
Time series	Show change over time	Line plot (`geom_line`)
Categorical + continuous	Compare groups	Bar chart (`geom_col`), Box plot (`geom_boxplot`)
1 continuous variable	Show distribution	Histogram (`geom_histogram`), Density plot
Multiple groups + continuous	Compare distributions	Box plot (`geom_boxplot`), Violin plot (`geom_violin`)
Counts of categories	Show frequencies	Bar chart (`geom_bar`)

5.15.1 Key Concepts Review

Three essentials: Data + aes() + geom_*()
Layers: Build plots by adding layers with +
Aesthetics: Map variables to visual properties (aes())
Geoms: Choose the right plot type for your data
Scales: Control how data maps to visuals
Labels: Always label axes and legends clearly (labs())
Themes: Customize appearance for publication
Save: Use ggsave() with appropriate format and resolution

5.16 Week 5 Homework: Creating Publication-Ready Plots

5.16.1 Assignment Overview

You will create 5 different types of plots from a provided dataset and present them in a professional Quarto report. Focus on choosing appropriate plot types, applying proper labels, and using themes effectively.

5.16.2 Dataset

You’ll use simulated data from a sheep grazing study:

# Generate dataset (include this in your homework)
set.seed(2024)

sheep_data <- tibble(
  sheep_id = rep(1:60, each = 4),
  breed = rep(rep(c("Merino", "Suffolk", "Dorper"), each = 20), each = 4),
  pasture_type = rep(rep(c("Native", "Improved"), each = 30), each = 4),
  week = rep(c(0, 4, 8, 12), times = 60),
  weight_kg = rep(rnorm(60, mean = 45, sd = 5), each = 4) +
              rep(c(0, 5, 9, 12), times = 60) +
              rnorm(240, mean = 0, sd = 1.5),
  wool_growth_mm = rep(rnorm(60, mean = 8, sd = 1.5), each = 4) +
                   rnorm(240, mean = 0, sd = 0.5)
)

# Add some additional variables
sheep_data <- sheep_data %>%
  group_by(sheep_id) %>%
  mutate(
    weight_gain_kg = weight_kg - first(weight_kg),
    avg_daily_gain_g = (weight_gain_kg * 1000) / (week * 7)
  ) %>%
  ungroup() %>%
  filter(week > 0)  # Remove week 0 for some analyses

5.16.3 Part 1: Scatter Plot (20 points)

Create a scatter plot showing the relationship between weight_kg (x-axis) and wool_growth_mm (y-axis).

Requirements: - Color points by breed - Set point size to 3 - Use theme_minimal() - Add appropriate title and axis labels - Include a subtitle indicating sample size

Bonus (+5 points): Add a trend line using geom_smooth(method = "lm") for each breed

5.16.4 Part 2: Line Plot (20 points)

Create a line plot showing weight_kg over week for individual sheep.

Requirements: - Show data for 6 randomly selected sheep (use filter(sheep_id %in% c(...))) - Color lines by breed - Add both lines and points - Use theme_classic() - Proper labels and title

5.16.5 Part 3: Bar Chart (20 points)

Create a bar chart showing the average final weight (week 12) by breed and pasture type.

Requirements: - Calculate average final weight using dplyr first - Use geom_col() with position = "dodge" for grouped bars - Fill bars by pasture_type - Add axis labels including units - Use a ColorBrewer palette

5.16.6 Part 4: Histogram (15 points)

Create histograms showing the distribution of avg_daily_gain_g for each breed.

Requirements: - Three separate histograms (one per breed) — hint: use faceting or create 3 separate plots - Choose appropriate binwidth - Fill color should differ by breed - Use theme_bw() - Title should be descriptive

Note: We’ll cover faceting (facet_wrap()) in the next chapter, but you can try it now or create 3 separate plots.

5.16.7 Part 5: Box Plot or Violin Plot (20 points)

Compare the distribution of weight_gain_kg (total gain over the study) across breeds and pasture types.

Requirements: - Use geom_boxplot() OR geom_violin() (your choice) - X-axis: breed, fill by pasture_type - Add jittered points with transparency - Use theme_light() - Clear labels and legend

5.16.8 Part 6: Interpretation (5 points)

For each plot, write 2-3 sentences interpreting what the plot shows. What patterns, trends, or differences do you observe?

5.16.9 Bonus Challenge (+10 points)

Create a single multi-panel figure combining your scatter plot and box plot using either: - Manual arrangement (creating two separate plots) - Looking ahead: Try library(patchwork) and combine with +

# Example structure (you'll learn this next week!)
library(patchwork)
plot1 + plot2

5.16.10 Submission Requirements

Submit a rendered Quarto HTML document
Include all code and output
All plots should be clearly labeled
Use professional formatting
Include interpretations

5.16.11 Recommended YAML

---
title: "Week 5 Homework: Data Visualization with ggplot2"
author: "Your Name"
date: today
format:
  html:
    toc: true
    toc-depth: 2
    code-fold: false
    theme: cosmo
    embed-resources: true
execute:
  warning: false
  message: false
---

5.16.12 Grading Rubric

Part 1: Scatter Plot (20%)
- Correct geom and aesthetics (8%)
- Proper labels and theme (7%)
- Overall appearance (5%)
Part 2: Line Plot (20%)
- Correct data filtering and plotting (8%)
- Lines and points shown (7%)
- Labels and styling (5%)
Part 3: Bar Chart (20%)
- Correct data summarization (8%)
- Grouped bars correctly displayed (7%)
- Labels and color palette (5%)
Part 4: Histogram (15%)
- Appropriate binwidth chosen (6%)
- One histogram per breed (5%)
- Labels and theme (4%)
Part 5: Box/Violin Plot (20%)
- Correct plot with multiple groupings (8%)
- Points added with transparency (7%)
- Labels and theme (5%)
Part 6: Interpretation (5%)
- Clear, accurate interpretations for each plot

5.17 Additional Resources

5.17.1 Required Reading

R for Data Science (2e) - Chapters 9-10: Visualize and Layers
ggplot2 documentation
ggplot2 book (3e) - Chapters 1-4

5.17.2 Optional Reading

Wilkinson, L. (2005). The Grammar of Graphics (2nd ed.). Springer. (Original theory)
Wickham, H. (2010). “A Layered Grammar of Graphics.” Journal of Computational and Graphical Statistics, 19(1), 3-28.
Data Visualization: A Practical Introduction by Kieran Healy

5.17.3 Videos

“ggplot2 Workshop” by Thomas Lin Pedersen (RStudio)
“A Quick Introduction to ggplot2” by StatQuest with Josh Starmer
“ggplot2 Tutorial” series by Data Science Dojo

5.17.4 Cheat Sheets

ggplot2 Cheat Sheet ⭐ Most important!
RStudio Cheatsheets Collection

5.17.5 Galleries and Inspiration

R Graph Gallery — Hundreds of examples with code
Top 50 ggplot2 Visualizations
ggplot2 extensions — Additional packages

5.17.6 Color Resources

ColorBrewer — Color schemes for maps and data viz
Viridis color palettes — Colorblind-friendly
Coolors — Color palette generator

5.17.7 Interactive Learning

5.17.8 Useful Websites

Next Chapter: Advanced ggplot2 and Multi-Panel Plots

# Introduction to Data Visualization with ggplot2 ## Learning Objectives By the end of this chapter, you will be able to: 1. Understand the Grammar of Graphics philosophy and how ggplot2 implements it 2. Create scatter plots, line plots, bar charts, histograms, box plots, and violin plots 3. Map variables to aesthetic properties (color, size, shape, alpha) 4. Apply and customize themes for publication-ready plots 5. Add labels, titles, and customize scales 6. Save plots with `ggsave()` for different output formats 7. Choose the appropriate plot type for your data and research question --- ## Introduction to the Grammar of Graphics ### What is ggplot2? `ggplot2` is R's most powerful and popular data visualization package. Created by Hadley Wickham, it implements Leland Wilkinson's **Grammar of Graphics** — a systematic way to describe and build any visualization. ::: {.callout-note} ## Why "gg"plot2? - **gg** = Grammar of Graphics - **2** = Second iteration (there was a ggplot1, but it's obsolete) The package is part of the tidyverse and integrates seamlessly with dplyr, tidyr, and other tidyverse tools. ::: ### The Grammar of Graphics Philosophy Traditional plotting systems (like base R graphics) force you to choose from predefined plot types. The Grammar of Graphics takes a different approach: **every visualization is built by combining independent components**. ```{mermaid} flowchart LR A[Data] --> B[Aesthetic Mappings] B --> C[Geometric Objects] C --> D[Statistical Transformations] D --> E[Scales] E --> F[Coordinate System] F --> G[Facets] G --> H[Theme] H --> I[Final Plot] ``` ### The Three Essential Components Every ggplot2 plot requires three components: 1. **Data**: A data frame or tibble 2. **Aesthetic mappings** (`aes()`): How variables map to visual properties (x, y, color, size, etc.) 3. **Geometric objects** (`geom_*()`): The type of plot (points, lines, bars, etc.) **Basic syntax**: ```r ggplot(data = <DATA>) + aes(x = <X_VAR>, y = <Y_VAR>) + geom_<TYPE>() ``` Or more commonly: ```r ggplot(<DATA>, aes(x = <X_VAR>, y = <Y_VAR>)) + geom_<TYPE>() ``` ::: {.callout-important} ## ggplot2 Uses `+`, Not `%>%` Notice we use `+` to add layers to a ggplot, NOT the pipe (`%>%` or `|>`). This is a common mistake! ```r # CORRECT ggplot(data, aes(x = var1, y = var2)) + geom_point() # WRONG - don't use pipe! ggplot(data, aes(x = var1, y = var2)) %>% geom_point() ``` ::: ### Setup Let's load the necessary packages: ```{r} #| eval: true #| echo: true #| message: false library(tidyverse) # Includes ggplot2, dplyr, etc. ``` --- ## Your First ggplot: Scatter Plots with `geom_point()` Scatter plots display the relationship between two continuous variables. Each observation is represented by a point. ### Basic Scatter Plot Let's create example data on cattle feed intake and weight gain: ```{r} #| eval: true #| echo: true # Create example cattle data cattle <- tibble( animal_id = paste0("C", 1001:1030), breed = rep(c("Holstein", "Jersey", "Angus"), each = 10), feed_intake_kg = c( rnorm(10, mean = 25, sd = 3), # Holstein rnorm(10, mean = 18, sd = 2), # Jersey rnorm(10, mean = 22, sd = 2.5) # Angus ), weight_gain_kg = c( rnorm(10, mean = 1.2, sd = 0.15), # Holstein rnorm(10, mean = 0.8, sd = 0.12), # Jersey rnorm(10, mean = 1.0, sd = 0.13) # Angus ) ) # View first few rows head(cattle) # Create basic scatter plot ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) + geom_point() ``` **Reading this code:** 1. `ggplot(cattle, ...)` — Use the cattle data 2. `aes(x = feed_intake_kg, y = weight_gain_kg)` — Map feed intake to x-axis, weight gain to y-axis 3. `geom_point()` — Draw points ### Adding Color by Group We can map the `breed` variable to color: ```{r} #| eval: true #| echo: true ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point() ``` Notice: - Each breed gets a different color automatically - A legend is created automatically - Colors are chosen from a default palette ### Customizing Point Aesthetics Points have several aesthetic properties we can map to variables OR set manually: | Aesthetic | What it controls | Example variable type | |-----------|------------------|-----------------------| | `x` | Horizontal position | Continuous | | `y` | Vertical position | Continuous | | `color` | Point outline color | Categorical or continuous | | `size` | Point size | Continuous | | `shape` | Point shape | Categorical | | `alpha` | Transparency (0-1) | Continuous | ::: {.callout-important} ## Mapping vs Setting **Mapping** (inside `aes()`): Connects a variable to an aesthetic — varies by data ```r aes(color = breed) # Color depends on breed variable ``` **Setting** (outside `aes()`): Sets a fixed value for all points ```r geom_point(color = "blue", size = 3) # All points blue and size 3 ``` ::: #### Example: Mapping Size ```{r} #| eval: true #| echo: true # Map size to feed intake (redundant with x, but demonstrates the concept) ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed, size = feed_intake_kg)) + geom_point() ``` #### Example: Setting Fixed Aesthetics ```{r} #| eval: true #| echo: true # Make all points larger and semi-transparent ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 4, alpha = 0.6) ``` #### Example: Using Shape ```{r} #| eval: true #| echo: true # Map shape to breed (in addition to color) ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed, shape = breed)) + geom_point(size = 3) ``` ### When to Use Scatter Plots ✅ **Use scatter plots when:** - You have two continuous variables - You want to explore relationships/correlations - You want to identify outliers or patterns - You want to show individual data points ❌ **Don't use scatter plots when:** - One or both variables are categorical (use bar charts or box plots) - You have too many overlapping points (consider hex bins or contour plots) - You want to show change over time for a single entity (use line plots) --- ## Line Plots with `geom_line()` Line plots connect points in order, typically used for time series or sequential data. ### Example: Growth Curves ```{r} #| eval: true #| echo: true # Create growth data for individual animals over time growth <- tibble( animal_id = rep(c("H001", "H002", "H003"), each = 6), week = rep(c(0, 2, 4, 6, 8, 10), times = 3), weight_kg = c( c(45, 58, 72, 88, 105, 122), # H001 c(42, 55, 68, 83, 98, 115), # H002 c(48, 60, 75, 91, 108, 126) # H003 ) ) growth # Create line plot ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) + geom_line() ``` ### Combining Lines and Points Often, it's helpful to show both the line and the data points: ```{r} #| eval: true #| echo: true ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) + geom_line() + geom_point() ``` ::: {.callout-tip} ## Layer Order Matters Layers are drawn in the order you add them. Later layers appear on top of earlier layers. ```r # Points on top of lines (usually preferred) geom_line() + geom_point() # Lines on top of points (can obscure data) geom_point() + geom_line() ``` ::: ### Customizing Lines ```{r} #| eval: true #| echo: true # Thicker lines with different line types ggplot(growth, aes(x = week, y = weight_kg, color = animal_id)) + geom_line(linewidth = 1.2) + geom_point(size = 3) ``` ### When to Use Line Plots ✅ **Use line plots when:** - Data represents a sequence (especially time) - You want to show trends or changes - Points should be connected in a specific order - You're tracking individual subjects over time ❌ **Don't use line plots when:** - Data points are independent (no meaningful order) - You have many overlapping trajectories (hard to read) - X-axis is not continuous or sequential --- ## Bar Charts with `geom_col()` and `geom_bar()` Bar charts display categorical data with rectangular bars. ### `geom_col()` vs `geom_bar()` | Function | When to use | Y-axis comes from | |----------|-------------|-------------------| | `geom_col()` | You've already calculated the values | Your data | | `geom_bar()` | You want to count occurrences | Automatic counting | ### Using `geom_col()` with Pre-calculated Values ```{r} #| eval: true #| echo: true # Average milk production by breed (already summarized) milk_production <- tibble( breed = c("Holstein", "Jersey", "Guernsey", "Brown Swiss"), avg_liters_per_day = c(35, 24, 28, 30) ) milk_production # Create bar chart ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) + geom_col() ``` ### Using `geom_bar()` to Count ```{r} #| eval: true #| echo: true # Create dataset with individual animals animals <- tibble( breed = sample(c("Holstein", "Jersey", "Angus"), size = 100, replace = TRUE) ) # Count animals per breed automatically ggplot(animals, aes(x = breed)) + geom_bar() ``` ::: {.callout-note} ## `geom_bar()` Counts for You When you use `geom_bar()`, you only specify `x` (the categorical variable). The y-axis (count) is calculated automatically using `stat = "count"`. ::: ### Adding Color ```{r} #| eval: true #| echo: true # Color bars by breed ggplot(milk_production, aes(x = breed, y = avg_liters_per_day, fill = breed)) + geom_col() ``` ::: {.callout-important} ## `color` vs `fill` for Bars - `color` controls the **outline** of the bar - `fill` controls the **inside** of the bar Usually, you want `fill` for bar charts: ```r aes(fill = breed) # Fills the bars (typical) aes(color = breed) # Just colors the outlines (usually not what you want) ``` ::: ### Grouped and Stacked Bars ```{r} #| eval: true #| echo: true # Create data with two categorical variables feed_trial <- tibble( breed = rep(c("Holstein", "Jersey"), each = 3), feed_type = rep(c("A", "B", "C"), times = 2), avg_gain_kg = c(1.2, 1.4, 1.1, 0.9, 1.0, 0.8) ) feed_trial # Grouped bars ggplot(feed_trial, aes(x = breed, y = avg_gain_kg, fill = feed_type)) + geom_col(position = "dodge") # Stacked bars (default) ggplot(feed_trial, aes(x = breed, y = avg_gain_kg, fill = feed_type)) + geom_col() ``` ### Horizontal Bars ```{r} #| eval: true #| echo: true # Use coord_flip() for horizontal bars ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) + geom_col() + coord_flip() ``` ### When to Use Bar Charts ✅ **Use bar charts when:** - Comparing quantities across categories - Showing counts or frequencies - Categories are on the x-axis - Bars should start at zero ❌ **Don't use bar charts when:** - Showing distributions (use histograms) - Comparing many categories (can be hard to read) - Changes over time are important (use line plots) --- ## Histograms with `geom_histogram()` Histograms show the **distribution** of a single continuous variable by dividing the data into bins and counting how many observations fall into each bin. ### Basic Histogram ```{r} #| eval: true #| echo: true # Create birth weight data birth_weights <- tibble( calf_id = 1:200, birth_weight_kg = rnorm(200, mean = 40, sd = 5) ) # Create histogram ggplot(birth_weights, aes(x = birth_weight_kg)) + geom_histogram() ``` ::: {.callout-warning} ## Pick the Right Number of Bins! The default number of bins (30) may not be ideal for your data. Experiment with: ```r geom_histogram(bins = 20) # Specify number of bins geom_histogram(binwidth = 2) # Specify width of each bin ``` Too few bins → lose detail; too many bins → noisy ::: ### Customizing Binwidth ```{r} #| eval: true #| echo: true # Wider bins (3 kg) ggplot(birth_weights, aes(x = birth_weight_kg)) + geom_histogram(binwidth = 3, fill = "steelblue", color = "white") ``` ### Overlapping Histograms by Group ```{r} #| eval: true #| echo: true # Create breed data with different distributions set.seed(123) breeds_vec <- sample(c("Holstein", "Jersey"), size = 200, replace = TRUE, prob = c(0.5, 0.5)) birth_weights_breed <- tibble( breed = breeds_vec, birth_weight_kg = ifelse( breed == "Holstein", rnorm(sum(breed == "Holstein"), mean = 42, sd = 5), rnorm(sum(breed == "Jersey"), mean = 37, sd = 4) ) ) # Overlapping histograms ggplot(birth_weights_breed, aes(x = birth_weight_kg, fill = breed)) + geom_histogram(binwidth = 2, alpha = 0.6, position = "identity") ``` ::: {.callout-tip} ## Use `alpha` for Overlapping Histograms When showing multiple distributions, use `alpha` (transparency) so you can see where they overlap: ```r geom_histogram(alpha = 0.5, position = "identity") ``` `position = "identity"` places histograms on top of each other (default is stacked). ::: ### When to Use Histograms ✅ **Use histograms when:** - Showing the distribution of a continuous variable - Checking for normality - Identifying skewness or outliers - Understanding the shape of your data ❌ **Don't use histograms when:** - You have categorical data (use bar charts) - Comparing multiple groups precisely (use box plots or density plots) - You have very few data points (distributions won't be clear) --- ## Box Plots with `geom_boxplot()` Box plots (box-and-whisker plots) display the distribution of data through five summary statistics: minimum, first quartile (Q1), median, third quartile (Q3), and maximum. ### Anatomy of a Box Plot ```{mermaid} graph TD A[Box Plot Components] --> B[Box: IQR Q1 to Q3] A --> C[Line in box: Median] A --> D[Whiskers: 1.5 × IQR] A --> E[Points beyond whiskers: Outliers] ``` ### Basic Box Plot ```{r} #| eval: true #| echo: true # Create data for different treatment groups treatments <- tibble( treatment = rep(c("Control", "Drug A", "Drug B", "Drug C"), each = 25), weight_gain_kg = c( rnorm(25, mean = 50, sd = 8), # Control rnorm(25, mean = 58, sd = 7), # Drug A rnorm(25, mean = 62, sd = 9), # Drug B rnorm(25, mean = 55, sd = 6) # Drug C ) ) # Create box plot ggplot(treatments, aes(x = treatment, y = weight_gain_kg)) + geom_boxplot() ``` ### Adding Color ```{r} #| eval: true #| echo: true ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) + geom_boxplot() ``` ### Combining Box Plots with Individual Points It's often useful to show the underlying data along with the box plot: ```{r} #| eval: true #| echo: true # Add jittered points ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) + geom_boxplot(alpha = 0.7) + geom_jitter(width = 0.2, alpha = 0.5) ``` ::: {.callout-tip} ## `geom_jitter()` Reduces Overplotting When points overlap, `geom_jitter()` adds small random horizontal noise to make them visible: ```r geom_jitter(width = 0.2) # Jitter horizontally (adjust width as needed) ``` ::: ### When to Use Box Plots ✅ **Use box plots when:** - Comparing distributions across groups - You want to show median, quartiles, and outliers - You have multiple groups to compare - You want a compact summary of distributions ❌ **Don't use box plots when:** - Your data has multiple modes (bimodal, multimodal) — box plots hide this - You have very few data points (< ~10 per group) - You want to show exact distributions (use violin plots or histograms) --- ## Violin Plots with `geom_violin()` Violin plots show the full distribution of data (like a density plot) rotated and mirrored. They're useful for showing distributions that box plots might hide. ### Basic Violin Plot ```{r} #| eval: true #| echo: true ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) + geom_violin() ``` ### Combining Violin Plots with Box Plots A powerful combination: violin plot shows distribution, box plot shows summary statistics: ```{r} #| eval: true #| echo: true ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) + geom_violin(alpha = 0.5) + geom_boxplot(width = 0.2, fill = "white") ``` ### Adding Points ```{r} #| eval: true #| echo: true ggplot(treatments, aes(x = treatment, y = weight_gain_kg, fill = treatment)) + geom_violin(alpha = 0.5) + geom_jitter(width = 0.1, alpha = 0.3, size = 1) ``` ### When to Use Violin Plots ✅ **Use violin plots when:** - You want to show full distribution shape - Your data might be bimodal or have multiple peaks - Comparing distributions across groups - You have enough data points (> ~20 per group) ❌ **Don't use violin plots when:** - You have very few data points (distributions won't be smooth) - Readers aren't familiar with violin plots (less intuitive than box plots) - You only need summary statistics (use box plots) --- ## Aesthetic Mappings: A Deeper Dive We've seen several aesthetics already. Let's consolidate what we know: ### Common Aesthetics | Aesthetic | Type | Example | Best used with | |-----------|------|---------|----------------| | `x` | Position | `aes(x = treatment)` | All geoms | | `y` | Position | `aes(y = weight)` | All geoms | | `color` | Color | `aes(color = breed)` | Points, lines, text | | `fill` | Fill | `aes(fill = breed)` | Bars, boxes, violins | | `size` | Size | `aes(size = age)` | Points, text | | `shape` | Shape | `aes(shape = sex)` | Points | | `alpha` | Transparency | `aes(alpha = density)` | All geoms | | `linetype` | Line type | `aes(linetype = group)` | Lines | ### Continuous vs Discrete Variables ggplot2 handles continuous and discrete variables differently: ```{r} #| eval: true #| echo: true # Create data with both variable types mixed_data <- tibble( group = rep(c("A", "B", "C"), each = 20), # Discrete score = rnorm(60, mean = 100, sd = 15), # Continuous intensity = rep(1:20, times = 3) # Continuous ) # Color by discrete variable (gets distinct colors) ggplot(mixed_data, aes(x = intensity, y = score, color = group)) + geom_point(size = 3) # Color by continuous variable (gets gradient) ggplot(mixed_data, aes(x = intensity, y = score, color = intensity)) + geom_point(size = 3) ``` ::: {.callout-important} ## Choosing the Right Aesthetic - **Categorical data**: Use `color`, `fill`, `shape`, `linetype` - **Continuous data**: Use `x`, `y`, `size`, `alpha`, `color` (gradient) - **Too many categories?** (> ~7): Don't use `shape` (hard to distinguish); consider faceting instead ::: --- ## Working with Scales Scales control how data values map to visual properties. Every aesthetic has a scale. ### Scale Functions Scale functions follow the pattern: `scale_<aesthetic>_<type>()` Examples: - `scale_x_continuous()` — Continuous x-axis - `scale_y_log10()` — Log10 transform y-axis - `scale_color_manual()` — Manually specify colors - `scale_fill_brewer()` — Use ColorBrewer palettes ### Modifying Continuous Scales ```{r} #| eval: true #| echo: true # Create sample data growth_rate <- tibble( animal_id = 1:30, weight_kg = runif(30, min = 400, max = 700), gain_per_day_kg = runif(30, min = 0.8, max = 1.5) ) # Modify axis limits and breaks ggplot(growth_rate, aes(x = weight_kg, y = gain_per_day_kg)) + geom_point() + scale_x_continuous(limits = c(350, 750), breaks = seq(350, 750, by = 100)) + scale_y_continuous(limits = c(0, 2), breaks = seq(0, 2, by = 0.5)) ``` ### Log Transformations ```{r} #| eval: true #| echo: true # Data with exponential relationship bacteria <- tibble( time_hours = 0:24, count = 100 * 2^(0:24 / 4) # Exponential growth ) # Regular scale (hard to see early values) ggplot(bacteria, aes(x = time_hours, y = count)) + geom_line() + geom_point() # Log scale (easier to see pattern) ggplot(bacteria, aes(x = time_hours, y = count)) + geom_line() + geom_point() + scale_y_log10() ``` ### Discrete Scales: Manual Colors ```{r} #| eval: true #| echo: true # Specify custom colors ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + scale_color_manual(values = c("Holstein" = "blue", "Jersey" = "orange", "Angus" = "darkred")) ``` ### Using ColorBrewer Palettes ```{r} #| eval: true #| echo: true # Use ColorBrewer palette (colorblind-friendly) ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + scale_color_brewer(palette = "Set1") ``` ::: {.callout-tip} ## Colorblind-Friendly Palettes Use `scale_color_viridis_d()` (discrete) or `scale_color_viridis_c()` (continuous) for accessible colors: ```r scale_color_viridis_d() # Discrete variable scale_color_viridis_c() # Continuous variable ``` ::: --- ## Labels and Titles with `labs()` Use `labs()` to add or modify plot labels: ```{r} #| eval: true #| echo: true ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + labs( title = "Relationship Between Feed Intake and Weight Gain", subtitle = "Data from 30 cattle across three breeds", x = "Daily Feed Intake (kg)", y = "Daily Weight Gain (kg)", color = "Breed", caption = "Source: Fictional farm data (2024)" ) ``` ::: {.callout-note} ## Label Arguments in `labs()` - `title` — Main plot title - `subtitle` — Secondary title (smaller font) - `x`, `y` — Axis labels - `color`, `fill`, `shape`, `size` — Legend titles (match the aesthetic name) - `caption` — Small text at bottom (often for data source or notes) ::: --- ## Themes: Customizing Plot Appearance Themes control the non-data elements of your plot (background, grid lines, fonts, etc.). ### Built-in Themes ggplot2 includes several complete themes: ```{r} #| eval: true #| echo: true # Create base plot p <- ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + labs(title = "Feed Intake vs Weight Gain") # Default theme (theme_gray) p # theme_minimal() - Clean and modern p + theme_minimal() # theme_classic() - No grid lines p + theme_classic() # theme_bw() - Black and white p + theme_bw() # theme_light() - Light gray grid p + theme_light() ``` ### Setting a Default Theme You can set a theme for all plots in your session: ```{r} #| eval: false #| echo: true # Set default theme for all subsequent plots theme_set(theme_minimal()) ``` ### Customizing Themes Use `theme()` to modify specific elements: ```{r} #| eval: true #| echo: true ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + labs(title = "Feed Intake vs Weight Gain") + theme_minimal() + theme( plot.title = element_text(size = 16, face = "bold"), axis.title = element_text(size = 12), legend.position = "bottom" ) ``` ::: {.callout-tip} ## Common Theme Customizations ```r theme( plot.title = element_text(size = 16, face = "bold"), axis.text = element_text(size = 10), legend.position = "bottom", # or "top", "left", "right", "none" panel.grid.major = element_line(color = "gray90"), panel.grid.minor = element_blank() ) ``` ::: --- ## Coordinate Systems Coordinate systems control how the x and y axes are rendered. ### `coord_cartesian()`: Zooming `coord_cartesian()` is useful for zooming without removing data: ```{r} #| eval: true #| echo: true # Zoom in on a specific region ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + coord_cartesian(xlim = c(18, 26), ylim = c(0.8, 1.2)) ``` ::: {.callout-important} ## `coord_cartesian()` vs `scale_*_continuous(limits = ...)` - `coord_cartesian(xlim = c(a, b))` — **Zooms** without removing data (stat calculations use all data) - `scale_x_continuous(limits = c(a, b))` — **Removes** data outside limits before plotting Usually, you want `coord_cartesian()` for zooming. ::: ### `coord_flip()`: Horizontal Plots We saw this earlier with bar charts. It swaps x and y: ```{r} #| eval: true #| echo: true ggplot(milk_production, aes(x = breed, y = avg_liters_per_day)) + geom_col(fill = "steelblue") + coord_flip() ``` ### Fixed Aspect Ratio `coord_fixed()` ensures equal scales on both axes (1 unit on x = 1 unit on y visually): ```{r} #| eval: true #| echo: true # Without fixed aspect ratio ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) + geom_point() # With fixed aspect ratio ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg)) + geom_point() + coord_fixed(ratio = 20) # 20 kg of feed = 1 kg weight gain visually ``` --- ## Saving Plots with `ggsave()` After creating a plot, save it with `ggsave()`: ```{r} #| eval: false #| echo: true # Create plot my_plot <- ggplot(cattle, aes(x = feed_intake_kg, y = weight_gain_kg, color = breed)) + geom_point(size = 3) + labs(title = "Feed Intake vs Weight Gain") + theme_minimal() # Save to file ggsave("cattle_plot.png", plot = my_plot, width = 8, height = 6, dpi = 300) ``` ### Common Arguments | Argument | What it does | Example | |----------|--------------|---------| | `filename` | File path and name | `"plots/figure1.png"` | | `plot` | ggplot object (defaults to last plot) | `my_plot` | | `width` | Width in inches (or cm, mm) | `8` | | `height` | Height in inches (or cm, mm) | `6` | | `dpi` | Resolution (dots per inch) | `300` (publication quality) | | `units` | Units for width/height | `"in"`, `"cm"`, `"mm"` | ### File Formats ggplot2 supports many formats (determined by file extension): - **PNG** (`.png`) — Good for presentations, web (raster) - **PDF** (`.pdf`) — Best for publications (vector, scalable) - **JPEG** (`.jpg`) — Compressed (raster, smaller files) - **SVG** (`.svg`) — Vector format for web - **TIFF** (`.tiff`) — High quality (raster, large files) ```{r} #| eval: false #| echo: true # Save as PDF (recommended for publications) ggsave("figure1.pdf", width = 7, height = 5) # Save as high-resolution PNG ggsave("figure1.png", width = 7, height = 5, dpi = 300) ``` ::: {.callout-tip} ## Publication Guidelines Most journals require: - **300 dpi** minimum resolution - **PDF** or **TIFF** format for final submission - Specific dimensions (check journal guidelines) - Readable fonts (usually at least 8-10 pt in final figure) Start with: ```r ggsave("figure.pdf", width = 7, height = 5, dpi = 300) ``` ::: --- ## Summary: Choosing the Right Plot Type | Data Type | Goal | Recommended Plot | |-----------|------|------------------| | 2 continuous variables | Show relationship | **Scatter plot** (`geom_point`) | | Time series | Show change over time | **Line plot** (`geom_line`) | | Categorical + continuous | Compare groups | **Bar chart** (`geom_col`), **Box plot** (`geom_boxplot`) | | 1 continuous variable | Show distribution | **Histogram** (`geom_histogram`), **Density plot** | | Multiple groups + continuous | Compare distributions | **Box plot** (`geom_boxplot`), **Violin plot** (`geom_violin`) | | Counts of categories | Show frequencies | **Bar chart** (`geom_bar`) | ### Key Concepts Review 1. **Three essentials**: Data + `aes()` + `geom_*()` 2. **Layers**: Build plots by adding layers with `+` 3. **Aesthetics**: Map variables to visual properties (`aes()`) 4. **Geoms**: Choose the right plot type for your data 5. **Scales**: Control how data maps to visuals 6. **Labels**: Always label axes and legends clearly (`labs()`) 7. **Themes**: Customize appearance for publication 8. **Save**: Use `ggsave()` with appropriate format and resolution --- ## Week 5 Homework: Creating Publication-Ready Plots ### Assignment Overview You will create **5 different types of plots** from a provided dataset and present them in a professional Quarto report. Focus on choosing appropriate plot types, applying proper labels, and using themes effectively. ### Dataset You'll use simulated data from a sheep grazing study: ```{r} #| eval: false #| echo: true # Generate dataset (include this in your homework) set.seed(2024) sheep_data <- tibble( sheep_id = rep(1:60, each = 4), breed = rep(rep(c("Merino", "Suffolk", "Dorper"), each = 20), each = 4), pasture_type = rep(rep(c("Native", "Improved"), each = 30), each = 4), week = rep(c(0, 4, 8, 12), times = 60), weight_kg = rep(rnorm(60, mean = 45, sd = 5), each = 4) + rep(c(0, 5, 9, 12), times = 60) + rnorm(240, mean = 0, sd = 1.5), wool_growth_mm = rep(rnorm(60, mean = 8, sd = 1.5), each = 4) + rnorm(240, mean = 0, sd = 0.5) ) # Add some additional variables sheep_data <- sheep_data %>% group_by(sheep_id) %>% mutate( weight_gain_kg = weight_kg - first(weight_kg), avg_daily_gain_g = (weight_gain_kg * 1000) / (week * 7) ) %>% ungroup() %>% filter(week > 0) # Remove week 0 for some analyses ``` ### Part 1: Scatter Plot (20 points) Create a scatter plot showing the relationship between `weight_kg` (x-axis) and `wool_growth_mm` (y-axis). **Requirements**: - Color points by `breed` - Set point size to 3 - Use `theme_minimal()` - Add appropriate title and axis labels - Include a subtitle indicating sample size **Bonus** (+5 points): Add a trend line using `geom_smooth(method = "lm")` for each breed ### Part 2: Line Plot (20 points) Create a line plot showing `weight_kg` over `week` for individual sheep. **Requirements**: - Show data for **6 randomly selected sheep** (use `filter(sheep_id %in% c(...))`) - Color lines by `breed` - Add both lines and points - Use `theme_classic()` - Proper labels and title ### Part 3: Bar Chart (20 points) Create a bar chart showing the **average final weight** (week 12) by breed and pasture type. **Requirements**: - Calculate average final weight using `dplyr` first - Use `geom_col()` with `position = "dodge"` for grouped bars - Fill bars by `pasture_type` - Add axis labels including units - Use a ColorBrewer palette ### Part 4: Histogram (15 points) Create histograms showing the distribution of `avg_daily_gain_g` for each breed. **Requirements**: - Three separate histograms (one per breed) — hint: use faceting or create 3 separate plots - Choose appropriate binwidth - Fill color should differ by breed - Use `theme_bw()` - Title should be descriptive **Note**: We'll cover faceting (`facet_wrap()`) in the next chapter, but you can try it now or create 3 separate plots. ### Part 5: Box Plot or Violin Plot (20 points) Compare the distribution of `weight_gain_kg` (total gain over the study) across breeds and pasture types. **Requirements**: - Use `geom_boxplot()` OR `geom_violin()` (your choice) - X-axis: `breed`, fill by `pasture_type` - Add jittered points with transparency - Use `theme_light()` - Clear labels and legend ### Part 6: Interpretation (5 points) For each plot, write 2-3 sentences interpreting what the plot shows. What patterns, trends, or differences do you observe? ### Bonus Challenge (+10 points) Create a single multi-panel figure combining your scatter plot and box plot using either: - Manual arrangement (creating two separate plots) - Looking ahead: Try `library(patchwork)` and combine with `+` ```{r} #| eval: false #| echo: true # Example structure (you'll learn this next week!) library(patchwork) plot1 + plot2 ``` ### Submission Requirements - Submit a rendered Quarto HTML document - Include all code and output - All plots should be clearly labeled - Use professional formatting - Include interpretations ### Recommended YAML ```yaml --- title: "Week 5 Homework: Data Visualization with ggplot2" author: "Your Name" date: today format: html: toc: true toc-depth: 2 code-fold: false theme: cosmo embed-resources: true execute: warning: false message: false --- ``` ### Grading Rubric - **Part 1: Scatter Plot (20%)** - Correct geom and aesthetics (8%) - Proper labels and theme (7%) - Overall appearance (5%) - **Part 2: Line Plot (20%)** - Correct data filtering and plotting (8%) - Lines and points shown (7%) - Labels and styling (5%) - **Part 3: Bar Chart (20%)** - Correct data summarization (8%) - Grouped bars correctly displayed (7%) - Labels and color palette (5%) - **Part 4: Histogram (15%)** - Appropriate binwidth chosen (6%) - One histogram per breed (5%) - Labels and theme (4%) - **Part 5: Box/Violin Plot (20%)** - Correct plot with multiple groupings (8%) - Points added with transparency (7%) - Labels and theme (5%) - **Part 6: Interpretation (5%)** - Clear, accurate interpretations for each plot --- ## Additional Resources ### Required Reading - [R for Data Science (2e)](https://r4ds.hadley.nz/) - Chapters 9-10: Visualize and Layers - [ggplot2 documentation](https://ggplot2.tidyverse.org/) - [ggplot2 book (3e)](https://ggplot2-book.org/) - Chapters 1-4 ### Optional Reading - Wilkinson, L. (2005). *The Grammar of Graphics* (2nd ed.). Springer. (Original theory) - Wickham, H. (2010). "A Layered Grammar of Graphics." *Journal of Computational and Graphical Statistics*, 19(1), 3-28. - [Data Visualization: A Practical Introduction](https://socviz.co/) by Kieran Healy ### Videos - "ggplot2 Workshop" by Thomas Lin Pedersen (RStudio) - "A Quick Introduction to ggplot2" by StatQuest with Josh Starmer - "ggplot2 Tutorial" series by Data Science Dojo ### Cheat Sheets - [ggplot2 Cheat Sheet](https://rstudio.github.io/cheatsheets/data-visualization.pdf) ⭐ **Most important!** - [RStudio Cheatsheets Collection](https://posit.co/resources/cheatsheets/) ### Galleries and Inspiration - [R Graph Gallery](https://r-graph-gallery.com/) — Hundreds of examples with code - [Top 50 ggplot2 Visualizations](http://r-statistics.co/Top50-Ggplot2-Visualizations-MasterList-R-Code.html) - [ggplot2 extensions](https://exts.ggplot2.tidyverse.org/) — Additional packages ### Color Resources - [ColorBrewer](https://colorbrewer2.org/) — Color schemes for maps and data viz - [Viridis color palettes](https://cran.r-project.org/web/packages/viridis/vignettes/intro-to-viridis.html) — Colorblind-friendly - [Coolors](https://coolors.co/) — Color palette generator ### Interactive Learning - [RStudio Primers: Visualize Data](https://posit.cloud/learn/primers/3) - [ggplot2 Exercises](https://r-graphics.org/exercises.html) ### Useful Websites - [ggplot2 Tidyverse Page](https://ggplot2.tidyverse.org/) - [Stack Overflow: ggplot2 tag](https://stackoverflow.com/questions/tagged/ggplot2) - [RStudio Community](https://community.rstudio.com/) --- **Next Chapter**: [Advanced ggplot2 and Multi-Panel Plots](ch06-ggplot2_advanced.qmd)

5.1 Learning Objectives

5.2 Introduction to the Grammar of Graphics

5.2.1 What is ggplot2?

5.2.2 The Grammar of Graphics Philosophy

5.2.3 The Three Essential Components

5.2.4 Setup

5.3 Your First ggplot: Scatter Plots with geom_point()

5.3.1 Basic Scatter Plot

5.3.2 Adding Color by Group

5.3.3 Customizing Point Aesthetics

5.3.3.1 Example: Mapping Size

5.3.3.2 Example: Setting Fixed Aesthetics

5.3.3.3 Example: Using Shape

5.3.4 When to Use Scatter Plots

5.4 Line Plots with geom_line()

5.4.1 Example: Growth Curves

5.4.2 Combining Lines and Points

5.4.3 Customizing Lines

5.4.4 When to Use Line Plots

5.5 Bar Charts with geom_col() and geom_bar()

5.5.1 geom_col() vs geom_bar()

5.5.2 Using geom_col() with Pre-calculated Values

5.5.3 Using geom_bar() to Count

5.5.4 Adding Color

5.5.5 Grouped and Stacked Bars

5.5.6 Horizontal Bars

5.5.7 When to Use Bar Charts

5.6 Histograms with geom_histogram()

5.6.1 Basic Histogram

5.6.2 Customizing Binwidth

5.6.3 Overlapping Histograms by Group

5.6.4 When to Use Histograms

5.7 Box Plots with geom_boxplot()

5.7.1 Anatomy of a Box Plot

5.7.2 Basic Box Plot

5.7.3 Adding Color

5.7.4 Combining Box Plots with Individual Points

5.7.5 When to Use Box Plots

5.8 Violin Plots with geom_violin()

5.8.1 Basic Violin Plot

5.8.2 Combining Violin Plots with Box Plots

5.8.3 Adding Points

5.8.4 When to Use Violin Plots

5.9 Aesthetic Mappings: A Deeper Dive

5.9.1 Common Aesthetics

5.9.2 Continuous vs Discrete Variables

5.10 Working with Scales

5.10.1 Scale Functions

5.10.2 Modifying Continuous Scales

5.10.3 Log Transformations

5.10.4 Discrete Scales: Manual Colors

5.10.5 Using ColorBrewer Palettes

5.11 Labels and Titles with labs()

5.12 Themes: Customizing Plot Appearance

5.12.1 Built-in Themes

5.12.2 Setting a Default Theme

5.12.3 Customizing Themes

5.13 Coordinate Systems

5.13.1 coord_cartesian(): Zooming

5.13.2 coord_flip(): Horizontal Plots

5.13.3 Fixed Aspect Ratio

5.14 Saving Plots with ggsave()

5.14.1 Common Arguments

5.14.2 File Formats

5.15 Summary: Choosing the Right Plot Type

5.15.1 Key Concepts Review

5.16 Week 5 Homework: Creating Publication-Ready Plots

5.16.1 Assignment Overview

5.16.2 Dataset

5.16.3 Part 1: Scatter Plot (20 points)

5.16.4 Part 2: Line Plot (20 points)

5.16.5 Part 3: Bar Chart (20 points)

5.16.6 Part 4: Histogram (15 points)

5.16.7 Part 5: Box Plot or Violin Plot (20 points)

5.16.8 Part 6: Interpretation (5 points)

5.16.9 Bonus Challenge (+10 points)

5.16.10 Submission Requirements

5.16.11 Recommended YAML

5.16.12 Grading Rubric

5.17 Additional Resources

5.3 Your First ggplot: Scatter Plots with `geom_point()`

5.4 Line Plots with `geom_line()`

5.5 Bar Charts with `geom_col()` and `geom_bar()`

5.5.1 `geom_col()` vs `geom_bar()`

5.5.2 Using `geom_col()` with Pre-calculated Values

5.5.3 Using `geom_bar()` to Count

5.6 Histograms with `geom_histogram()`

5.7 Box Plots with `geom_boxplot()`

5.8 Violin Plots with `geom_violin()`

5.11 Labels and Titles with `labs()`

5.13.1 `coord_cartesian()`: Zooming

5.13.2 `coord_flip()`: Horizontal Plots

5.14 Saving Plots with `ggsave()`