Scatter Plots: Visualizing Relationships Between Variables
In the world of data science and statistics, seeing is believing. A scatter plot—also known as a scattergram, scatter graph, or scatter diagram—is a powerful mathematical tool used to display the relationship between two variables using Cartesian coordinates. By plotting individual data points on a horizontal and vertical axis, researchers can instantly identify patterns, trends, and correlations that might remain hidden in a raw spreadsheet.
Beyond simple two-dimensional charts, scatter plots can be enhanced by coding points with different colors, shapes, or sizes to represent a third variable. This makes them an essential component of the seven basic tools of quality control, helping professionals monitor processes and identify deviations.
Key Facts
- Purpose: To visualize the relationship or correlation between two continuous variables.
- Axes: The independent variable (control parameter) is typically placed on the horizontal x-axis, while the dependent variable is placed on the vertical y-axis.
- Correlation Types: Patterns can show positive (upward sloping), negative (downward sloping), or null (no relationship) correlations.
- Advanced Forms: Can be expanded into bubble charts, marker charts, or complex scatter plot matrices.
- Statistical Utility: Used to determine lines of best fit through methods like linear regression.
The History of Scatter Plot Development
The evolution of the scatter plot distinguishes it from standard line charts. While line charts connect data points to show change over time, scatter plots represent specific observations of bivariate data (data involving two variables).
The origins of these visualizations are debated. While Edmund Halley created a bivariate plot of temperature and pressure in 1686, he did not include the specific data points. Most historians, including Michael Friendly and Daniel Denis, attribute the first true scatter plot to John Herschel in 1833. Herschel used freehand drawing to plot the angle of a star in the constellation Virgo over time.
Later, Sir Francis Galton expanded the utility of the scatter plot. In 1886, Galton published a scatter plot regarding the height of parents and children, introducing the concept of binning and averaging adjacent cells to create smoother, more readable visualizations. This work laid the foundation for statisticians like Karl Pearson and R.A. Fisher to formalize correlation and significance testing.
Understanding Correlations and Trends
When analyzing a scatter plot, the primary goal is often to identify the degree of correlation between variables. It is important to note that correlation does not imply causation; a scatter plot shows how variables move together, not necessarily that one causes the other.
Types of Correlation Patterns
- Positive Correlation: The dots form a pattern that slopes from the lower left to the upper right, indicating that as one variable increases, the other also increases.
- Negative Correlation: The dots slope from the upper left to the lower right, indicating that as one variable increases, the other decreases.
- Null Correlation: The dots appear scattered randomly with no discernible slope, suggesting no relationship exists.
To better understand these relationships, statisticians often draw a line of best fit, also known as a trendline. For linear relationships, a procedure called linear regression is used to calculate this line. For non-linear relationships, smooth lines such as LOESS can be applied to visualize complex patterns.
In complex datasets, a 3D scatter plot can be used to visualize multivariate data. By using different axes for different scalar variables, researchers can map coordinates in a phase space, often using color or glyphs to represent even more dimensions of data.
![A 3D scatter plot allows the visualization of multivariate data. This scatter plot takes multiple scalar variables and uses them for different axes in phase space. The different variables are combined to form coordinates in the phase space and they are displayed using glyphs and coloured using another scalar variable.[6]](/images/84/eb/84eb64dabda47c48efae613bfddec52066e9a39d1a0a8f1918d3103e602afd43.jpg)
Practical Application: A Lung Capacity Example
To visualize how lung capacity affects breath-holding duration, a researcher would plot each participant's data as a single coordinate. For instance, a person with a lung capacity of 400 cl who can hold their breath for 21.7 seconds would be represented by the point (400, 21.7) on the Cartesian plane. By plotting an entire group, the researcher can visually confirm if a relationship exists between these two physiological measurements.
Scatter Plot Matrices
When dealing with a large set of variables (X₁, X₂, ... Xₖ), a single plot is insufficient. A scatter plot matrix solves this by displaying all possible pairwise scatter plots in a single grid. For a dataset with *k* variables, the matrix will consist of *k* rows and *k* columns. The cell at the intersection of the *i*-th row and *j*-th column shows the relationship between variable Xᵢ and Xⱼ.
Generalized matrices can also handle combinations of categorical and quantitative variables, using specialized displays like mosaic plots or faceted bar charts to provide a comprehensive view of the entire dataset.
Summary of Scatter Plot Characteristics
| Feature | Description |
|---|---|
| X-Axis | Typically represents the independent or control variable. |
| Y-Axis | Typically represents the dependent or measured variable. |
| Data Points | Individual observations represented as dots or markers. |
| Trendline | A line of best fit used to model the relationship (e.g., linear regression). |
| Dimensions | Can be 2D, 3D, or multivariate using color/size coding. |
Frequently Asked Questions
What is the difference between a scatter plot and a line chart?
A line chart is typically used to show how a single variable changes over a continuous interval (like time), connecting points with lines. A scatter plot displays individual observations of two different variables to reveal the relationship or correlation between them.
Can a scatter plot show more than two variables?
Yes. While a standard plot uses two axes, you can add a third variable by using color, shape, or the size of the points (often called a bubble chart). 3D scatter plots can also be used to add a third physical axis.
Does a strong correlation in a scatter plot prove causation?
No. A scatter plot can show that two variables are highly correlated (they move together), but it does not prove that one variable causes the change in the other.
What is a line of best fit?
A line of best fit, or trendline, is a straight or curved line drawn through the data points that best represents the overall direction of the data. It is often calculated using mathematical procedures like linear regression.
What is a scatter plot matrix?
A scatter plot matrix is a grid of multiple scatter plots that allows you to see the pairwise relationships between all variables in a large dataset simultaneously.