1 Frequency Polygon Basics

1.1 Definition and purpose

A frequency polygon is a line graph used to represent the distribution of numerical data across categories. For grouped data, it is typically made by plotting the midpoint of each class interval on the horizontal axis and the corresponding class frequency (or relative frequency) on the vertical axis. The plotted points are then joined with straight line segments, producing a polygonal line that summarizes how often values fall within each interval.

The main purpose of a frequency polygon is to provide a compact visual summary of the overall pattern of a dataset—such as whether observations concentrate in the middle of the range or spread toward one side. Because it emphasizes the changing frequency from one interval to the next, it is well suited for qualitative description and for comparing distributions.

1.2 Relationship to histograms

A frequency polygon is closely related to a histogram. Both are based on the same underlying class intervals and class counts. In a histogram, the frequencies are represented by contiguous bars, whereas in a frequency polygon the information is encoded by discrete points connected by lines.

As a result, a frequency polygon can often be viewed as a “line version” of a histogram: it preserves the general shape and level changes across classes, but it avoids the need to draw filled rectangles. This difference can make the line easier to overlay when comparing two or more groups.

1.3 When to use frequency polygons

Frequency polygons are commonly used when:

  • The data are already summarized in class intervals.
  • Comparing distributions across groups is a priority.
  • A quick visual judgment about skewness or modality is desired.
  • One wants a plot that remains readable when multiple distributions are drawn on the same axes.

They are frequently used in introductory statistics because the construction is straightforward and the graph communicates distribution shape effectively.

2 Constructing a Frequency Polygon

2.1 Data requirements

2.1.1 Grouped vs. ungrouped data

Frequency polygons are most direct for grouped data, where each class interval has an associated frequency. If the data are ungrouped, they can be converted into grouped form by choosing class intervals and counting observations within each interval, after which the frequency polygon can be constructed in the standard way.

Grouping decisions affect the appearance of the resulting polygon, so class interval choices should be made consistently with the context of the analysis.

2.1.1.1 Class intervals and frequencies

To build a frequency polygon from grouped data, one needs:

  • A set of class intervals (e.g., 10–19, 20–29, 30–39).
  • The frequency in each interval (the number of observations falling into that class).
  • A clear convention for the horizontal axis scale, typically using class midpoints.

Once these elements are available, the polygon can be drawn without using the raw dataset directly.

2.2 Determining class midpoints

For each class interval, the class midpoint is computed as the average of the class boundaries. For example, for an interval with lower boundary \(L\) and upper boundary \(U\), the midpoint is \[ M=\frac{L+U}{2}. \] Midpoints provide a representative x-value for each class, allowing the distribution to be plotted as a sequence of connected points.

When working with integer-valued classes (such as 10–19), it is common to interpret the class boundaries as 9.5–19.5 (so the midpoint becomes 14.5). The boundary convention should match the method used to define the histogram or frequency table.

2.3 Plotting points and connecting them

The standard construction steps are:

  1. Compute a midpoint for each class interval.
  2. Determine the corresponding frequency (or relative frequency).
  3. Plot each pair \((\text{midpoint}, \text{frequency})\) on a coordinate system.
  4. Connect neighboring points with straight line segments.

The resulting line represents how the frequency changes from one class to the next. Labeling axes clearly is important, including units for the variable and the meaning of the vertical scale.

2.4 Including boundary classes (if used)

Some presentations add extra “boundary” points beyond the first and last class to close the shape visually. Typically, this is done by creating two additional points at midpoints one class width beyond the ends, with a frequency of zero. These added points can make the polygon appear more like a complete curve and can improve interpretability when comparing graphs.

Whether to include boundary classes is a stylistic choice; it does not change the internal class information, but it can affect the visual framing.

3 Interpreting a Frequency Polygon

3.1 Shape descriptors (symmetry, skewness, modality)

A frequency polygon is often interpreted by describing its overall shape:

  • Symmetry: If the left and right sides mirror each other in rise and fall, the distribution is roughly symmetric.
  • Skewness: If the peak is closer to one side and the tail extends farther on the other side, the distribution is skewed (right-skewed if values extend toward higher intervals; left-skewed otherwise).
  • Modality: The number of prominent peaks indicates the number of modes. A single peak suggests a unimodal distribution, while two separated peaks suggest bimodality.

Because the polygon connects class midpoints, the shape is most reliable when class intervals are reasonably consistent in width and when there are enough observations to reduce random fluctuation.

3.2 Comparing multiple distributions

A major advantage of frequency polygons is their compatibility with overlays. Multiple polygons can be plotted on the same axes using different colors or line styles to compare patterns across groups.

When comparing, analysts typically look for:

  • Differences in peak heights (indicating relative concentration).
  • Shifts in where the peak occurs (indicating changes in central tendency).
  • Differences in spread and tail behavior (indicating variability and skewness).

If vertical scales represent different types of frequencies (e.g., counts for one plot and relative frequencies for another), comparisons should be treated cautiously.

3.3 Reading approximate values from the graph

A frequency polygon allows approximate reading of:

  • Where concentrations occur: by locating peaks and troughs along the horizontal axis.
  • How frequencies change: by comparing vertical heights between adjacent points.
  • Approximate magnitudes: by referencing the axis ticks.

However, because the polygon is constructed from grouped intervals, it reflects interval-level information rather than exact individual values. Fine-grained conclusions about observations within an interval are generally not supported by the polygon alone.

4 Frequency Polygon Variations

4.1 Relative frequency polygon

In a relative frequency polygon, the vertical axis uses relative frequencies rather than raw counts. Relative frequency is commonly computed as \[ \text{relative frequency}=\frac{f}{N}, \] where \(f\) is the class frequency and \(N\) is the total number of observations.

This variation is especially useful when comparing groups of different sizes. By converting frequencies to proportions, the polygons become comparable in height and shape without being dominated by sample size.

4.2 Cumulative frequency polygon (contrast)

A cumulative frequency polygon uses cumulative frequencies, where each point represents the total frequency up to and including a given class (or sometimes up to the previous class, depending on convention). This is related to cumulative distribution ideas.

In contrast to the standard frequency polygon, which focuses on how frequencies are distributed across intervals, the cumulative version emphasizes accumulation and is often used to study how data gather as the variable increases. It can be useful for reading approximate quantiles when combined with appropriate axis interpretation.

4.3 Smoothed/approximate polygons (practical use)

In some practical settings, analysts may smooth the line to reduce visual noise from sampling variability or from sparse class counts. Smoothing typically involves replacing sharp point-to-point changes with a gentler curve or applying a moving average approach.

While smoothing can make large-scale trends more apparent, it also introduces additional modeling choices and can obscure genuine features such as small secondary peaks. Any smoothed polygon should be clearly labeled so that readers understand it is an approximation.

5 Common Pitfalls and Best Practices

5.1 Incorrect midpoint placement

A frequent error is using incorrect midpoints, such as averaging class labels rather than class boundaries under a consistent boundary convention. Another issue occurs when class intervals have been defined with inclusive or exclusive rules that are not aligned with how midpoints are calculated.

Best practice is to compute midpoints from the actual class boundaries (or from consistently defined endpoints) and to document the convention used.

5.2 Misaligned frequency scale

If the vertical axis is mislabeled or scaled inconsistently—such as mixing counts and relative frequencies or using different scaling factors across groups—then the plotted heights can mislead interpretation.

To avoid this, one should:

  • Verify whether the y-axis represents frequency or relative frequency.
  • Use the same vertical scaling when overlaying multiple polygons.
  • Ensure axis units and labels match the plotted quantities.

5.3 Inconsistent class intervals

Frequency polygons assume a coherent class-interval structure. If one group uses different class widths or different interval boundaries, the resulting polygons may not be directly comparable.

A best practice is to standardize class intervals across groups, or if standardization is not possible, to consider re-binning or using methods designed for differing group summaries.

5.4 Misleading comparisons due to scaling

Even with correct data, comparisons can be misleading if axes are rescaled in a way that changes perceived differences. For example, changing y-axis limits or using different scales across separate graphs can exaggerate or understate contrast.

When comparing distributions visually, it is generally preferable to maintain consistent axis ranges and scaling across the figures, especially when the intent is side-by-side comparison.

6 Worked Examples

6.1 Example with grouped data

Suppose a dataset is grouped into class intervals of width 10:

  • 0–9: frequency 4
  • 10–19: frequency 8
  • 20–29: frequency 13
  • 30–39: frequency 9
  • 40–49: frequency 6

Assume boundaries align with these intervals so midpoints are:

  • 0–9 midpoint 4.5
  • 10–19 midpoint 14.5
  • 20–29 midpoint 24.5
  • 30–39 midpoint 34.5
  • 40–49 midpoint 44.5

To draw the frequency polygon, plot the points:

  • (4.5, 4)
  • (14.5, 8)
  • (24.5, 13)
  • (34.5, 9)
  • (44.5, 6)

Connect them in order with straight segments. The peak occurs at the midpoint 24.5, indicating the highest concentration of observations in the 20–29 interval.

6.2 Example comparing two groups

Consider two groups, A and B, both summarized with the same class intervals (0–9, 10–19, 20–29, 30–39, 40–49).

Frequencies for group A:

  • 0–9: 2
  • 10–19: 6
  • 20–29: 11
  • 30–39: 8
  • 40–49: 3

Frequencies for group B:

  • 0–9: 5
  • 10–19: 9
  • 20–29: 7
  • 30–39: 4
  • 40–49: 1

Compute the same midpoints as before (4.5, 14.5, 24.5, 34.5, 44.5). Plot and connect points for each group using different line styles.

Interpretation:

  • Group A peaks around 20–29, indicating higher central concentration in that range.
  • Group B has a comparatively higher early-class frequency (0–9 and 10–19) and declines more rapidly, suggesting a distribution shifted toward lower values.

6.3 Example using relative frequencies

Using the same group A frequencies from the previous example:

  • 2, 6, 11, 8, 3

Total \(N = 2+6+11+8+3 = 30\). Relative frequencies are:

  • 0–9: \(2/30 \approx 0.067\)
  • 10–19: \(6/30 = 0.200\)
  • 20–29: \(11/30 \approx 0.367\)
  • 30–39: \(8/30 \approx 0.267\)
  • 40–49: \(3/30 = 0.100\)

Plot points:

  • (4.5, 0.067)
  • (14.5, 0.200)
  • (24.5, 0.367)
  • (34.5, 0.267)
  • (44.5, 0.100)

Connect them to form the relative frequency polygon. Compared with the count-based polygon, the shape will be comparable even if another group had a different total sample size.