2.6.6. Geoms in ggplot2

1. Introduction

A "geom" in ggplot2 is the geometric object that represents your data in a plot. Each plot type uses a different geom, such as points, bars, or boxes. Understanding geoms is key to creating a wide variety of visualizations in R, including for clinical datasets such as SDTM or ADaM.

2. What is a Geom?

• A geom is the shape or object used to represent data in a plot.
• Examples: points (scatterplots), bars (barplots), rectangles (histograms), boxes (boxplots).
• You choose the geom by adding a layer like geom_point(), geom_bar(), or geom_boxplot() to your ggplot2 code.

3. Histograms: geom_histogram

• Purpose: Visualize the distribution and shape of a single numeric variable (e.g., AGE or HEIGHT in a clinical dataset).
• When to use: To see the range, center, and spread of your data.

R Code:

library(ggplot2)
# Example SDTM/ADaM-like dataset
adam <- data.frame(
  USUBJID = paste0("SUBJ", 1:100),
  AGE = sample(18:80, 100, replace = TRUE),
  HEIGHT = round(rnorm(100, mean = 165, sd = 10), 1),
  WEIGHT = round(rnorm(100, mean = 70, sd = 12), 1),
  SEX = sample(c("M", "F"), 100, replace = TRUE),
  TRT = sample(c("Placebo", "Active"), 100, replace = TRUE)
)
ggplot(data = adam) +
  geom_histogram(mapping = aes(AGE), bins = 15)

• By default, uses 30 bins. Here, bins are set to 15 for clarity.

Expected Outcome:

A histogram showing the age distribution of subjects in the clinical dataset.

4. Barplots: geom_bar

• Purpose: Show counts for each category of a categorical variable (e.g., treatment arm or sex).
• When to use: To compare the size of groups.

R Code:

ggplot(data = adam) +
  geom_bar(mapping = aes(TRT))

Expected Outcome:

A barplot showing the number of subjects in each treatment arm.

5. Boxplots: geom_boxplot

• Purpose: Summarize the distribution of a numeric variable by category (e.g., weight by treatment arm).
• When to use: To compare medians, ranges, and outliers between groups.

R Code:

ggplot(data = adam) +
  geom_boxplot(mapping = aes(x = TRT, y = WEIGHT))

Expected Outcome:

A boxplot showing the weight distribution for each treatment arm.

6. Customizing Geoms

• You can change the appearance of geoms by adding arguments inside or outside aes().
• For example, change the fill color of all boxes in a boxplot:

R Code:

ggplot(data = adam) +
  geom_boxplot(mapping = aes(x = TRT, y = WEIGHT), fill = "lightblue")

Expected Outcome:

A boxplot with all boxes filled in light blue.

7. Exploring Other Geoms

• ggplot2 supports many geoms: geom_line(), geom_violin(), geom_density(), geom_area(), geom_jitter(), geom_smooth(), and more.
• For example, a line plot of mean weight by age group:

R Code:

library(dplyr)
adam$AGEGRP <- cut(adam$AGE, breaks = c(18,30,40,50,60,80), right = FALSE)
mean_weight <- adam %>%
  group_by(AGEGRP) %>%
  summarise(mean_weight = mean(WEIGHT, na.rm = TRUE))
ggplot(mean_weight, aes(x = AGEGRP, y = mean_weight, group = 1)) +
  geom_line() +
  geom_point()

8. Input and Output Table for Geoms

9. Exploring Beyond Basic Geoms

• Combine multiple geoms in one plot (e.g., points and smooth lines).
• Use facet_wrap() to split plots by category (e.g., by SEX).
• Customize geoms with color, size, linetype, and more.
• Explore advanced geoms in the ggplot2 documentation.

10. Practice Problems

1. Create a histogram of subject height.
2. Make a barplot of subject sex.
3. Draw a boxplot of weight by sex.
4. Change the fill color of a boxplot.
5. Combine a scatterplot and a smooth line for age vs. weight.

11. Further Reading and Resources

• ggplot2 documentation
• R Graph Gallery: ggplot2 Geoms
• R for Data Science: Data Visualization

**Resource download links**

2.6.6.-Geoms-in-ggplot2.zip