How to Compare Box Plots (With Examples)

Deconstructing the Box Plot: The Five Number Summary

The foundation of every box plot rests on the five number summary. These five values provide a comprehensive statistical snapshot of the dataset, marking the boundaries and center of the distribution. When visualizing these statistics, we create a structure that immediately communicates the data’s central location and variability.

A box plot is a type of plot that displays the five number summary of a dataset, which includes:

• The minimum value: The smallest observation in the dataset, excluding outliers.
• The first quartile (Q1): Also known as the 25th percentile, marking the boundary below which 25% of the data falls.
• The median (Q2): The central value, or the 50th percentile, which divides the dataset into two equal halves.
• The third quartile (Q3): Also known as the 75th percentile, marking the boundary below which 75% of the data falls.
• The maximum value: The largest observation in the dataset, excluding outliers.

The calculation of these five values is essential because they define the structure of the visualization. The differences in these summary statistics between datasets are precisely what we analyze when performing a comparative analysis, offering insight into how data distributions differ across groups.

Visualizing Data Distribution: Anatomy of Whiskers and Boxes

To construct a box plot, the central “box” is drawn spanning from Q1 to Q3. This box encapsulates the middle 50% of the data, providing a direct visual representation of the density and spread of the most common values. The length of this box is equivalent to the Interquartile Range (IQR), which is a crucial measure of dispersion.

To make a box plot, we draw a box from the first to the third quartile. Then we draw a vertical line at the median. Lastly, we draw “whiskers” from the quartiles to the minimum and maximum value (or to the calculated inner fences, depending on the software used). This structure quickly communicates the data’s central location and variability.

A vertical line is drawn inside the box to designate the median (Q2). This line’s position relative to Q1 and Q3 helps analysts quickly assess the internal distribution and potential skewness of the data. The entire structure allows for an intuitive interpretation of the entire dataset’s spread and central location.

The Core Methodology for Comparative Analysis

When presented with multiple box plots, the goal is to systematically compare the underlying statistical distributions they represent. This comparative analysis moves beyond merely looking at means and standard deviations, offering a richer, non-parametric view of data behavior. By addressing four primary areas of comparison, we can formulate robust conclusions about how different groups interact or perform relative to one another.

A rigorous comparison requires focusing on four critical elements displayed by the plots: central tendency, variability, shape, and extreme values. Each element provides a piece of the puzzle, contributing to a holistic understanding of the datasets. Ignoring any one of these aspects risks drawing incomplete or misleading conclusions about the populations being studied.

When comparing two or more box plots, we can answer four different questions to achieve a comprehensive analysis:

Criterion 1: Comparing Central Tendency (The Median)

The most immediate comparison involves the central location of the data, which is best represented by the median (Q2). Since the median is the line inside the box, comparing the vertical position of this line across different plots tells us which dataset generally contains higher or lower values. This statistic is preferred over the mean for box plot comparison because it is robust against the distorting influence of extreme values.

1. How do the median values compare? We compare the vertical line in each box to determine which dataset has a higher median value. If the median line in Plot A sits higher on the axis than the median line in Plot B, then, on average, the typical values in Dataset A are greater than those in Dataset B. This is a vital first step in determining performance differences between groups.

Furthermore, observing the overlap, or lack thereof, between the central boxes (the IQR) can give preliminary insights into the practical significance of the median difference. A large vertical separation between the medians, especially when the boxes themselves do not overlap much, suggests a meaningful difference in central location that warrants closer attention.

Criterion 2: Assessing Data Spread (The Interquartile Range)

The variability or spread of the data is measured most effectively using the Interquartile Range (IQR), which is the range covered by the box itself (Q3 minus Q1). The IQR represents the middle 50% of the data points and is an excellent measure of data dispersion because it is resistant to the influence of extreme values and outliers.

2. How does the dispersion compare? We can compare the length of each box (which represents the distance between Q1 and Q3 – the Interquartile Range (IQR)) to determine which dataset is more spread out. A longer box indicates greater variability in the middle half of the data, suggesting that the scores or measurements within that group are less consistent or more heterogeneous.

The overall range (from the end of the lower whisker to the end of the upper whisker) offers a broader view of the spread, encompassing all non-outlier data points. However, the IQR remains the primary indicator of typical variability, providing context regarding the reliability and consistency of the data distribution. A smaller IQR is generally indicative of a more homogeneous set of observations.

Criterion 3: Identifying Distribution Shape (Skewness)

Skewness describes the asymmetry in the probability distribution of a dataset. Box plots are excellent for quickly visualizing skewness by examining the positioning of the median line within the box and the relative lengths of the whiskers. A perfectly symmetrical distribution would have the median centered exactly within the box, and the two whiskers would be roughly equal in length.

3. How does the skewness compare? The relative position of the median line within the box is key. The closer the vertical line is to Q1, the more positively skewed (or right-skewed) the dataset. This implies that the distribution has a longer tail extending toward higher values. The closer the vertical line is to Q3, the more negatively skewed (or left-skewed) the dataset, indicating a concentration of data at the higher end and a longer tail extending toward lower values.

Comparing the skew across multiple plots is essential for understanding underlying distribution characteristics that might not be captured by central tendency alone. For example, if one distribution is negatively skewed (like Study Method 1 in the example), it suggests that most participants achieved high scores, whereas a symmetrical distribution might imply a more even spread across all scores.

Criterion 4: Detecting Anomalies (Outliers)

Outliers are extreme observations that deviate significantly from other observations in the dataset. While they can sometimes represent errors in measurement, they often signal important anomalies or unique events within the data that warrant further investigation. Box plots offer a standardized method for identifying these extreme values visually.

4. Are outliers present? In box plots, outliers are typically represented by tiny circles that extend beyond either whisker. The whisker length is capped based on the Interquartile Range (IQR) to ensure that any data point exceeding this boundary is mathematically classified as an outlier. An observation is defined to be an outlier if it meets one of the following criteria:

• An observation is less than Q1 – 1.5*IQR (Lower Fence)
• An observation is greater than Q3 + 1.5*IQR (Upper Fence)

Comparing the presence and magnitude of outliers between groups is critical for data quality assessment. If one group consistently displays more high-side outliers, it might indicate occasional superior performance or unique, non-generalizable factors influencing that group.

Case Study: Comparing Study Methods

The following example demonstrates how to apply these four comparative criteria. Consider two experimental datasets displaying the exam scores for students who used one of two studying techniques to prepare for the exam. We aim to determine which method yielded superior and more consistent results.

The datasets, consisting of twenty scores each, are provided below:

Method 1: 78, 78, 79, 80, 80, 82, 82, 83, 83, 86, 86, 86, 86, 87, 87, 87, 88, 88, 88, 91

Method 2: 66, 66, 66, 67, 68, 70, 72, 75, 75, 78, 82, 83, 86, 88, 89, 90, 93, 94, 95, 98

If we create box plots for each dataset, here’s what the visual comparison reveals:

Detailed Analysis of Study Method Performance

We can systematically compare these two box plots by addressing the four key questions:

1. How do the median values compare? The line in the middle of the box plot for Study Method 1 is significantly higher than the line for Study Method 2 (approximately 86 vs. 78). This outcome clearly indicates that the students who used Study Method 1 had a higher typical (median) exam score.

2. How does the dispersion compare? The box plot for Study Method 2 is visually much longer than Study Method 1. This means the Interquartile Range is greater for Method 2, indicating that the exam scores are much more spread out among students who used Study Method 2. Study Method 1 yielded more consistent results.

3. How does the skewness compare? The line in the middle of the box plot for Study Method 1 is close to Q3, which indicates that the distribution of exam scores for students who used Study Method 1 is negatively skewed. This means the majority of scores are clustered toward the higher end. Conversely, the median line in the box plot for Study Method 2 is near the center of the box, which means the distribution of scores has very little skewness, approaching symmetry.

4. Are outliers present? Neither box plot has tiny circles or markers that extend beyond the top or bottom whiskers. This absence confirms that neither dataset contained any clear statistical outliers, ensuring that the range and quartiles accurately reflect the entire sample distribution.

Synthesizing Insights from Comparative Box Plots

The ability of box plots to synthesize complex distributional characteristics into a simple, easy-to-read graph makes them invaluable for data presentation and comparison. Based on the higher median score and significantly smaller IQR, Study Method 1 resulted in both higher average performance and more reliable, consistent results among students, despite its slight negative skew.

Mastering the interpretation of the four comparative criteria—median, IQR, skewness, and outliers—equips any data professional with the tools necessary to translate raw data into clear, narrative insights that drive decision-making. These plots are a powerful starting point for any statistical investigation involving group comparisons.