Skip to content
MathClicksIGCSE← All notes Cambridge IGCSE® Mathematics 0580 Extended Chapter 20 — Histograms and Frequency Distribution Diagrams

20.2  Cumulative Frequency

Cumulative frequency curves, and the median, quartiles, interquartile range and percentiles you read from them — plotted at upper bounds, always.

Building a cumulative frequency table

A grouped frequency table tells you how many values land in each class. A cumulative frequency table answers a different question: how many values are less than or equal to each class boundary. You build it with a running total — add the frequencies up as you go down the table.

Key idea Cumulative frequency = frequency so far. Each cumulative value counts everything up to and including that class, so it belongs to the class’s upper boundary: “46 students took 40 seconds or less”. The final cumulative frequency must equal n, the total number of values — if it does not, an addition slipped somewhere.

Here is the data this whole page will use — the times, t seconds, taken by 80 students to solve a puzzle:

Time (seconds)FrequencyCumulative frequencyPoint to plot
0 < t ≤ 2066(20, 6)
20 < t ≤ 30146 + 14 = 20(30, 20)
30 < t ≤ 402620 + 26 = 46(40, 46)
40 < t ≤ 502046 + 20 = 66(50, 66)
50 < t ≤ 701466 + 14 = 80(70, 80)
Watch out — upper bounds, alwaysEvery point is plotted at the upper class boundary, because that is the value the running total belongs to. Plotting at midpoints (that is the frequency polygon habit) or at lower bounds is the classic error on this topic, and it drags every reading — median, quartiles, everything — off target. The very first cross of all goes at the lower bound of the first class at height 0: here (0, 0), because nobody took 0 seconds or less.
Example 1The heights, h cm, of 60 seedlings: 0 < h ≤ 10 (4 seedlings), 10 < h ≤ 20 (12), 20 < h ≤ 30 (22), 30 < h ≤ 40 (14), 40 < h ≤ 50 (8). Construct the cumulative frequency table, state the points to plot, and say where the curve will rise most steeply.
1
Run the total downwards, one addition per row. 4
4 + 12 = 16
16 + 22 = 38
38 + 14 = 52
52 + 8 = 60
2
Check the last value against the total: 60 seedlings ✓. Each cumulative frequency is paired with its upper boundary. plot (0, 0), (10, 4), (20, 16), (30, 38), (40, 52), (50, 60)
3
The curve climbs fastest where the most data is packed — the class with the largest frequency. steepest between h = 20 and h = 30 (22 seedlings, the biggest jump: 16 → 38)
cf = 4, 16, 38, 52, 60 — plotted at 10, 20, 30, 40, 50 — steepest over 20–30
Exam tip When a Paper 4 table has a cumulative frequency column with gaps, fill in every entry before touching the graph. Each value is one addition — and the final entry must be n. That single check catches almost every slip.

Plotting the cumulative frequency curve

Now plot the points and join them. The syllabus is precise about how: mark each plotted point clearly with a small cross (×), then join the crosses with a smooth curve — not dot-to-dot straight lines. The result is an elongated S-shape (sometimes called an ogive): flat where data is scarce, steep where data is crowded.

  1. Horizontal axis: the variable, on a proper continuous scale with units.
  2. Vertical axis: cumulative frequency, from 0 up to at least n.
  3. Start at the lower bound of the first class at height 0 — here (0, 0).
  4. Cross at every upper boundary: (20, 6), (30, 20), (40, 46), (50, 66), (70, 80).
  5. Join with a single smooth, always-rising curve.

Cumulative frequency curve — times of 80 students

0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 Time (seconds) Cumulative frequency steepest here — the crowded classes
Why the S-shape worksThe gradient of the curve is the local crowding of the data — in fact it is the frequency density from 20.1 wearing a different hat. Few students finished before 20 seconds, so the curve starts shallow; the 30–50 second classes hold 46 of the 80 students, so the curve races upwards there; the last few slow solvers flatten it out again. A cumulative frequency curve can never dip downwards — a running total cannot shrink. If yours does, a point is misplotted.
Exam tipCambridge mark schemes award the curve marks for crosses within half a small square of the correct positions, joined by a smooth increasing curve. Sharpen the pencil, plot the crosses first, check them against the table, and only then draw the curve through them in one motion.

Median, quartiles and the interquartile range

The whole point of the curve is that you can now read off measures of average and spread without a single calculation on the raw data. Everything is done with the dashed-line method: pick a position on the cumulative frequency axis, rule a dashed line across to the curve, then drop down and read the value on the horizontal axis.

median at n2    lower quartile at n4    upper quartile at 3n4 Reading positions on the cumulative frequency axis — curve convention

For our 80 students: read the median at 80 ÷ 2 = 40, the lower quartile (LQ) at 80 ÷ 4 = 20 and the upper quartile (UQ) at 3 × 80 ÷ 4 = 60. On a curve you use n/2, not the (n + 1)/2 rule from listed data — with a large grouped data set the difference is invisible at graph accuracy, and n/2 is the convention mark schemes quote.

The dashed-line method on the same curve

0 20 40 60 80 0 20 40 60 LQ 30 med ≈38 UQ 47 Time (seconds) Cumulative frequency

Across from 20 → down to LQ = 30 s. Across from 40 → down to median ≈ 38 s. Across from 60 → down to UQ = 47 s. Always draw the dashed lines on the exam paper — they are part of the method mark.

Key idea — the interquartile range IQR = UQ − LQ — here 47 − 30 = 17 seconds. It is the width of the interval holding the middle half (middle 50%) of the data. Because it ignores the bottom quarter and the top quarter completely, one freakishly slow solver cannot move it — unlike the range, the IQR is immune to extreme values. Small IQR = consistent data; large IQR = spread-out data.

Percentiles

Quartiles are just special percentiles: LQ is the 25th, the median the 50th, UQ the 75th. For the p-th percentile, read across from p% of n. The 90th percentile of our times sits at position 0.9 × 80 = 72 on the cumulative frequency axis — 90% of students finished at or below that time.

Quartile reader

Switch data set to change n, and watch the reading positions n/4, n/2, 3n/4 move and the dashed lines re-draw on the curve. Then dial in any percentile of your own.

Example 2Use the curve above to estimate the median, the quartiles and the interquartile range of the 80 puzzle times, showing the method.
1
Work out the three reading positions first, before touching the graph. n = 80:  LQ at 80 ÷ 4 = 20,  median at 80 ÷ 2 = 40,  UQ at 3 × 80 ÷ 4 = 60
2
Dash across from 20 on the cumulative frequency axis to the curve, then down. LQ = 30 seconds
3
Repeat from 40 and from 60. median ≈ 38 seconds    UQ = 47 seconds
4
Subtract for the interquartile range — one single number. IQR = 47 − 30 = 17 seconds
LQ = 30 s,  median ≈ 38 s,  UQ = 47 s,  IQR = 17 s
Why “estimate”? The raw times were grouped before the curve was drawn, so any value read from it is an estimate — and mark schemes accept a small tolerance (typically half a small square). Say “estimate” in your answer; never claim graph readings are exact.

Reverse readings and comparing distributions

So far we started on the cumulative frequency axis. Exam questions love the other direction too: start from a value, go up to the curve, and read off how many. And because the curve counts “less than or equal to”, questions about more than need one extra subtraction.

“How many took 35 s or less?”

Up from t = 35 to the curve, across to the axis: about 32. Read the answer straight off.

“How many took more than 35 s?”

Same reading, then subtract from the total: 80 − 32 = 48. Forgetting the subtraction is the classic slip — the curve never counts “more than” directly.

A count between two values is a double read: cumulative frequency at the upper value minus cumulative frequency at the lower value.

Comparing two curves

Given two cumulative frequency curves on the same axes (two classes, two brands, two seasons), the exam wants two comparison sentences, each quoting numbers, each in context:

The box plot — the curve’s summary

The five numbers you can read from a cumulative frequency curve — minimum, LQ, median, UQ, maximum — are exactly the five numbers a box-and-whisker plot displays. For our puzzle data (fastest solver 8 s, slowest 68 s):

Box plot of the 80 puzzle times, on the same scale

min 8 LQ 30 median 38 UQ 47 max 68 0 10 20 30 40 50 60 70 Time (seconds)

The box spans LQ to UQ, so the box width is the IQR (17 s) and half of all students are inside it; the whiskers reach the minimum and maximum. It is the same information as the curve, compressed into five numbers.

Exam tipIn a comparison answer, vague wins nothing: “B is better” scores 0. The mark scheme wants a comparative word (higher, smaller, more consistent), the numbers quoted, and the context named (scores, times, masses). One sentence for medians, one for IQRs — that is the whole recipe.

More worked examples

Example 3Using the curve of the 80 puzzle times, the reading at t = 55 is a cumulative frequency of 70. Estimate (a) how many students took more than 55 seconds, (b) how many took between 30 and 55 seconds.
1
(a) The reading 70 counts students at 55 s or less. “More than” is everyone else — subtract from the total. more than 55 s:  80 − 70 = 10 students
2
(b) A between-count is the difference of two readings. From the table, the cumulative frequency at 30 s is exactly 20. between 30 and 55:  70 − 20 = 50 students
3
Sanity check: 20 at or below 30 s, 50 between, 10 above 55 s — and 20 + 50 + 10 = 80 ✓
(a) 10 students    (b) about 50 students
Watch out Never answer a “more than” question with the raw reading. The cumulative frequency axis always counts up to a value — the subtraction n − reading is where the mark lives.
Example 4For the same 80 students, estimate the 90th percentile of the times, and interpret it.
1
Convert the percentile into a position on the cumulative frequency axis. position = 90% of 80 = 0.9 × 80 = 72
2
Dash across from 72 to the curve and drop down. The curve gives a reading of about 59. 90th percentile ≈ 59 seconds
3
Interpret in context, both ways round: 90% of the students solved the puzzle in 59 seconds or less — equivalently, only 10% (8 students) took longer than about 59 s.
90th percentile ≈ 59 s — 90% of students finished within 59 s
Example 5Write down the five-figure summary of the puzzle times (fastest 8 s, slowest 68 s) and use it to find the range and the IQR. Which of the two is the better measure of spread here, and why?
1
Collect the five numbers — two extremes from the data, three readings from the curve. min 8,  LQ 30,  median 38,  UQ 47,  max 68
2
Range uses the extremes; IQR uses the quartiles. range = 68 − 8 = 60 s
IQR = 47 − 30 = 17 s
3
The range depends entirely on the single fastest and single slowest student, so one unusual solver distorts it. The IQR measures the middle 50% and ignores both tails, so it is the more reliable measure of how spread out a typical student’s time is.
range = 60 s,  IQR = 17 s — IQR is the better measure (not affected by extremes)
Example 6Two classes sat the same test (marks out of 100). From their cumulative frequency curves: Class A has LQ 45, median 58, UQ 74; Class B has LQ 60, median 66, UQ 78. Compare the two classes’ performances.
1
Compute the one number the question does not hand you: each IQR. IQR(A) = 74 − 45 = 29     IQR(B) = 78 − 60 = 18
2
Average sentence, with numbers and context: Class B’s median mark (66) is higher than Class A’s (58),
so Class B did better on average.
3
Spread sentence, with numbers and context: Class B’s IQR (18) is smaller than Class A’s (29),
so Class B’s marks are more consistent.
B: higher median (66 > 58) and smaller IQR (18 < 29) — better and more consistent
Exam trap The IQR is a single number, not an interval — writing “Class B’s IQR is 60 to 78” loses the mark. And a comparison without context (“B’s numbers are bigger”) scores nothing: name the medians, name the IQRs, and say what they mean about the marks.

Your turn

Practice set

  1. A table has frequencies 5, 9, 16, 7, 3 for classes with upper bounds 10, 20, 30, 40, 50. Write down the cumulative frequencies and the five points to plot.
  2. A cumulative frequency curve is drawn for n = 60 values. At which cumulative frequencies do you read the LQ, the median and the UQ?
  3. From a curve, the readings are LQ = 24, median = 32 and UQ = 45. Find the interquartile range and state what it measures.
  4. A curve shows the masses of 200 parcels. The reading at 50 kg is 158. Estimate how many parcels have a mass greater than 50 kg.
  5. Two shops’ daily takings: Shop A median $62, IQR $10; Shop B median $55, IQR $26. Write two sentences comparing the shops.

  1. Running total: 5, 14, 30, 37, 40 — plot at the upper bounds: (10, 5), (20, 14), (30, 30), (40, 37), (50, 40) (and start the curve at cumulative frequency 0 on the first lower bound).
  2. LQ at 60 ÷ 4 = 15, median at 60 ÷ 2 = 30, UQ at 3 × 60 ÷ 4 = 45.
  3. IQR = 45 − 24 = 21 — a single number: the width of the interval containing the middle half of the data.
  4. 158 parcels are at 50 kg or less, so heavier than 50 kg: 200 − 158 = 42 parcels.
  5. Shop A’s median takings ($62) are higher than Shop B’s ($55), so A takes more on a typical day. Shop A’s IQR ($10) is smaller than Shop B’s ($26), so A’s takings are more consistent.

Key points to remember

  • Cumulative frequency is a running total of the frequencies — the last value must equal n, and that is your built-in check.
  • Plot every point at the upper class boundary — never at midpoints or lower bounds — mark points with small crosses (×) and join them with a smooth increasing curve.
  • Read the median at n/2, the LQ at n/4 and the UQ at 3n/4 on the cumulative frequency axis, using dashed lines across then down.
  • IQR = UQ − LQ, a single number: the spread of the middle half of the data, immune to extreme values.
  • The p-th percentile is read at p% of n; the quartiles are the 25th, 50th and 75th percentiles.
  • “More than x” = n − reading at x; a between-count is the difference of two readings. Every value from a curve is an estimate.
  • Comparing two data sets: one sentence on the medians (average), one on the IQRs (consistency) — quote the numbers and name the context.

Where next?

✅ Mark scheme All notes

Progress is saved on this device only and appears on the notes home page.