Scatterplots

Scatterplots are one of the most fundamental tools for visualizing the relationship between two continuous variables (or variables that we wish to treat as continuous). Each point on the plot represents one observational unit’s values on the two plotted variables. This makes scatterplots particularly useful for identifying trends, clusters, and extreme values in the data.

Do: We make a simple scatterplot with the plot() function. For example, the following code uses the diabetes dataset to visualize the bivariate association between Body Mass Index, bmi, and Total Cholesterol, tc. Higher body fat is generally linked to less favorable cholesterol levels in the clinical literature.

plot(y = diabetes$tc, x = diabetes$bmi)

See: every point represents a different patient’s value of BMI and their level of total cholesterol. The points form a cloud with a slight upward trend, suggesting that higher BMI values tend to be associated with higher cholesterol levels, although the variability is substantial.

We can improve the readability by adding axis labels, a title, and adjusting the axis limits. To do so we adjust the following parameters.

Predict: these arguments only change the plot’s cosmetics; before running the code, predict whether the underlying pattern of points (their positions relative to each other) will change at all.

plot(y = diabetes$tc,
     x = diabetes$bmi,
     ylab = "Total Cholesterol",
     xlab = "Body Mass Index",
     main = "Relation between BMI and Cholesterol",
     ylim = c(0, 350),
     xlim = c(0, 50))

Explain: as predicted, the points themselves sit in exactly the same relative positions as before; only the labels, title, and the visible range of each axis have changed, making the figure easier to read without altering the underlying data.

When we apply plot() to a data frame containing only numeric columns, we generate a scatterplot matrix: a grid of scatterplots showing all the pairwise relationships. Do: Run the code below that selects 5 numeric columns.

plot(diabetes[c("age", "bmi", "tc", "glu", "progress")])

See: The grid contains 5 × 5 = 25 panels, one for every pair of the 5 selected variables (including each variable plotted against itself along the diagonal). Reading across a row or down a column lets us compare one variable’s relationship with every other variable at a glance, without having to build each scatterplot individually.

Practice Exercise

Using the diabetes dataset, make a scatterplot of total cholesterol on the y-axis against age on the x-axis. Add meaningful axis labels and a title.

  • The diabetes dataset is already loaded in the working directory of this webr session.
plot(y = diabetes$tc,
     x = diabetes$age,
     ylab = "Total Cholesterol",
     xlab = "Age",
     main = "Relation between Age and Cholesterol")

Back to top