The ggplot2 package creates plots by combining data, mappings, and layers. A simple bar chart of enrollment by site can be created as follows:

library(tidyverse)
prepared_data |>
ggplot(aes(x = site)) +
geom_bar() +
labs(
title = “Enrollment by Site”,
x = “Site”,
y = “Number of participants”
)

The ggplot() function defines the dataset and aesthetic mapping. The aes(x = site)
mapping places site on the x-axis. The geom_bar() layer counts records in each site category. The labs() function adds readable labels. This plot is useful only if each row represents one participant. If the dataset has repeated visits, the chart may overcount participants.
For a participant-level enrollment summary from a dataset with repeated rows, create the summary first:

site_enrollment <- prepared_data |>
distinct(participant_id, site) |>
count(site, name = “participants”)
site_enrollment |>
ggplot(aes(x = reorder(site, participants), y = participants)) +
geom_col() +
145
coord_flip() +
labs(
title = “Participants Enrolled by Site”,
x = “Site”,
y = “Participants”
)

The distinct() function ensures that each participant is counted once per site. The
geom_col() layer uses precomputed counts. coord_flip() makes long site names easier to read. The reorder() function orders sites by enrollment count.
A histogram can show the distribution of age:

prepared_data |>
ggplot(aes(x = age_years_derived)) +
geom_histogram(binwidth = 5, boundary = 0, color = “white”) +
labs(
title = “Age Distribution at Enrollment”,
x = “Age, years”,
y = “Number of participants”
)

The choice of bin width affects interpretation. A very small bin width may create a noisy chart, while a very large bin width may hide meaningful patterns. The data manager should experiment but avoid manipulating bins to exaggerate a point.
Box plots are useful for comparing numeric distributions across groups:

prepared_data |>
ggplot(aes(x = site, y = age_years_derived)) +
geom_boxplot() +
coord_flip() +
labs(
title = “Age Distribution by Site”,
x = “Site”,
y = “Age, years”
)

This plot can reveal differences in age distribution by site, but it should be interpreted with sample size. A site with five participants may have a box plot that looks unstable. It may be useful to add counts in a companion table.

The best way to learn visualization is to create simple plots, inspect them, and ask whether they answer the intended question.