Frequency Polygons

A histogram draws a solid bar over each class of grouped data. A frequency polygon tells the same story with a single wiggly line instead. The recipe is simple:

That joined-up line is the frequency polygon. Because it is just a line, it shows the overall shape of a distribution at a glance — where the data pile up, and how the tails trail off — and, best of all, you can lay two polygons on the same axes to compare distributions, something a pair of histograms does clumsily.

a fish a longer fish

Picture a histogram of fish lengths. Put a dot at the middle of the top of each bar — right above the class midpoint, at the bar's height — and connect the dots. You have drawn a frequency polygon straight from the histogram. The polygon is the histogram's "skyline", so a tall bar becomes a high point on the line and a short bar a low one.

Building one from a table

A grower weighs the apples off a tree and groups the weights into five 10-gram classes. To draw the polygon we only need two columns: the midpoint of each class, and its frequency.

Class (g) Midpoint x Frequency f
0–1052
10–20155
20–30259
30–40356
40–50453

The midpoint of 010 is \frac{0 + 10}{2} = 5, of 1020 is 15, and so on — the very same class midpoints used to find the estimated mean. That gives five points to plot, (5, 2), (15, 5), (25, 9), (35, 6) and (45, 3) — then we join them up. Press Play to add the points one class at a time:

The line climbs to a peak over the 2030 class (midpoint 25) and then falls away — most apples weigh about 25 grams, with fewer very light or very heavy ones.

Comparing two distributions

The real power of a frequency polygon shows when you draw two on the same axes. Here are the apples from Tree A (the table above) and a second Tree B, whose apples run lighter:

Both trees produced 25 apples, so the polygons are directly comparable. Tree B's line peaks further left, over the lighter 1020 class, while Tree A peaks over 2030. In one glance you can see that Tree A grows heavier apples — a comparison that two separate histograms would make you work for.

The traps that catch people out with frequency polygons:

an owl a weighing scale

Two schools measure their pupils' heights and group them into 10-centimetre bands. Overlay the two frequency polygons and the answer to "which school is taller on average?" jumps out: whichever line's peak sits further to the right along the height axis. The polygon turns a comparison of two long tables into a single picture your eye can read in a second.

Read a frequency polygon

Below is a frequency polygon for five equal classes, drawn with its histogram bars faint behind it so you can see the points sit on the midpoint of each bar's top. Press Refresh for fresh data, then practise: