G3 SEC Mathematics · Learner’s Guide Vol 0103. A clear route from class frequencies to cumulative totals, quartile estimates, grouped-data questions and worked answers.
A cumulative-frequency graph often looks easy to read. The line climbs from left to right, and the vertical axis counts more and more people or observations. Yet one small misunderstanding can spoil several questions: a point on the graph normally tells you how many observations have accumulated by a boundary, not how many sit inside the latest interval. To find an interval frequency, you usually have to subtract two cumulative counts.
This G3 SEC Mathematics learner’s guide teaches the complete route from an ordinary frequency table to cumulative totals, interval counts, medians, quartiles, percentiles and fair comparisons. We use original numerical examples, an independent practice set, worked corrections and a short revision sequence. The examples are written to teach the method, not to reproduce or predict an examination paper.
The starting point is Volume 0099 on graph scales. That workshop showed why the scale can change the apparent size of a difference. Here we focus on what a cumulative-frequency value actually counts. Use the Mathematics Hub if you need to repair earlier graph or fraction foundations before continuing.
Chapter index
- 1. The quick answer: cumulative means counted so far
- 2. Check the official SEC K310 scope
- 3. Build the cumulative table from ordinary frequencies
- 4. Boundaries and the meaning of “more than” or “at most”
- 5. Read a cumulative-frequency curve
- 6. How to find the median in grouped data
- 7. Lower quartile, upper quartile and interquartile range
- 8. Percentiles: a position is not the measured score
- 9. Comparing two distributions of different sizes
- 10. Common errors and their direct repairs
- 11. Original practice questions, from basic to advanced
- 12. Answer clinic with explicit reasoning
- 13. A worked study routine for grouped distributions
- 14. A seven-day progression with independent checks
- 15. The final link: the graph is a count of a set
1. The quick answer: cumulative means counted so far
Suppose a club records the number of students who completed a reading challenge within specified time limits. Four students completed by 10 minutes, nine by 20 minutes, fifteen by 30 minutes and twenty by 40 minutes. Those are cumulative totals. At the 30-minute boundary, the number fifteen includes the students already counted at 10 and 20 minutes.
How many completed after 20 minutes but by 30 minutes? Take the cumulative total by 30 and subtract the total by 20: 15−9=6. Do not answer fifteen. That would count everyone who finished by 30, including those who finished earlier.
The difference between a cumulative count and an interval count is the first idea you must secure. A cumulative count answers “how many up to this boundary?” An interval count answers “how many between these two boundaries?” The second uses a difference of two cumulative values.
A simple way to say it is: the upper boundary gives you the larger group, the lower boundary gives you the part you have already included, and subtraction leaves the interval. That is the same reasoning used when finding a segment of distance from two cumulative odometer readings. It is not a new trick unique to statistics.
If a question asks for the total population, read the final cumulative frequency when the table or graph covers all observations. If it asks for how many have values above a given boundary, subtract the cumulative count at that boundary from the total. You are now using the complement of the group already counted.
2. Check the official SEC K310 scope
The official 2027 SEC G3 Mathematics syllabus K310 includes the interpretation of cumulative-frequency diagrams, quartiles and percentiles, and measures of central tendency and spread. It also covers statistical representations such as histograms with equal class intervals and box-and-whisker plots. This chapter focuses on grouped data and cumulative interpretation, not every topic in Statistics and Probability.
The SEAB 2027 school-candidate syllabus directory lists Mathematics as K310, replacing the earlier reference code 4052. For the actual examination, use the syllabus for the year and level at which you are sitting the paper rather than assuming that every previous examination format carries forward unchanged.
K310 Paper 1 lasts 2 hours 15 minutes, has about 26 compulsory short-answer questions and carries 90 marks, worth half of the assessment. Paper 2 also lasts 2 hours 15 minutes, has 9 to 10 compulsory questions, carries 90 marks and contributes the other half. Its last question is a real-world-context application. Essential working matters; omitting it can result in lost marks.
Unless the question requires another accuracy, non-exact numerical answers are generally to three significant figures, and angles in degrees to one decimal place. A numerical estimate read from a cumulative-frequency diagram should follow any stated accuracy instruction. The way the graph has been drawn may also limit how precisely a value can reasonably be read.
The syllabus supplies the assessment frame. It does not specify that a particular example about reading challenges or travel times will appear. All tasks here are original practice designed to build a capability that applies when data are presented in unfamiliar contexts.
3. Build the cumulative table from ordinary frequencies
Here is an original dataset of 40 observations grouped into five non-overlapping intervals. Four fall in the interval above 0 and up to 20. Nine fall above 20 and up to 40. Fifteen fall above 40 and up to 60. Eight fall above 60 and up to 80. Four fall above 80 and up to 100. The frequencies are 4, 9, 15, 8 and 4.
The total is 4+9+15+8+4=40. That sum is your first check. If a question says there are 40 observations but your intervals sum to 38, investigate before drawing any graph. Missing counts in the starting table will affect every cumulative value.
To make the cumulative totals, start at the lower end and keep adding. Up to 20 there are 4 observations. Up to 40 there are 4+9=13. Up to 60 there are 13+15=28. Up to 80 there are 28+8=36. Up to 100 there are 36+4=40. The cumulative list is 4, 13, 28, 36, 40.
The final cumulative total is 40, agreeing with the ordinary-frequency sum. That check catches mistakes before they travel into graph drawing, quartile reading or interval interpretation. The cumulative numbers must not decrease as the boundary increases, because they represent a growing set of included observations.
Now reverse the operation. The number above 40 and up to 60 is 28−13=15. The number above 60 and up to 80 is 36−28=8. The number above 80 and up to 100 is 40−36=4. You recover the ordinary frequencies by subtracting consecutive cumulative totals.
For a table whose first boundary is zero, the cumulative count at that starting boundary is usually zero when the intervals and conventions specify that no observations are included there. When data contain observations exactly on a boundary, however, the definition of the intervals matters. In a real problem, use the stated convention rather than assume which side of a boundary owns the value.
4. Boundaries and the meaning of “more than” or “at most”
Many mistakes arise from treating different inequalities as the same sentence. “At most 40” includes values equal to 40. “Less than 40” excludes values equal to 40. “More than 40” excludes 40. “At least 40” includes it. In continuous grouped data, a graph’s class-boundary convention helps decide which readings are included.
Our illustrative intervals are described as above the lower boundary and up to the upper boundary. This means a value exactly 40 belongs to the interval above 20 and up to 40, not the next interval above 40 and up to 60. The two intervals do not overlap. That prevents double-counting when we subtract cumulative totals.
Suppose the cumulative table tells us 13 observations are at most 40 and 28 are at most 60. The number greater than 40 but at most 60 is 28−13=15. If the question instead asks for values from 40 inclusive to 60 inclusive, the table may not contain enough detail about how many observations equal 40. Do not pretend the grouped counts give an exact answer to a differently defined interval unless the conventions settle the issue.
This distinction matters in reading comprehension as well as mathematics. A statement such as “students scoring above 60” has a different boundary from “students scoring 60 or above”. Before calculating, underline the comparison words and map them to the actual intervals.
In an examination task, you may be able to read a cumulative-frequency curve at a specified continuous boundary and subtract as directed. When the data are discrete scores, be careful that a grouped display may summarise rather than list individual values. The numerical answer is only as exact as the data and grouping permit.
A dependable working line names the interval: “Number with 40 < score ≤ 60 = CF(60)−CF(40)=28−13=15.” This is clearer than writing only “28−13” with no explanation of what the numbers mean.
5. Read a cumulative-frequency curve
A cumulative-frequency curve typically has the variable or class boundary on the horizontal axis and cumulative frequency on the vertical axis. The points represent accumulated counts at those boundaries. The graph rises or remains level; it should not fall when all observations are counted once and the accumulation direction is increasing.
From our 40-observation table, the plotted points include (20,4), (40,13), (60,28), (80,36) and (100,40), with a suitable starting point at the lower boundary according to the dataset. Joining points appropriately produces a curve that summarises the grouped distribution. The curve’s steepness across a class interval reflects how many observations accumulate across that interval relative to the interval width.
Notice the difference between the y-coordinate and an interval frequency. At (60,28), the graph says 28 observations have values at or below 60 under the stated convention. It does not say 28 observations lie between 40 and 60. The class frequency is the vertical difference 28−13=15 over that interval.
If the graph is hand-drawn, values read between plotted points are estimates. A smooth interpolation does not reconstruct the exact positions of unseen individual observations. When a question asks for an estimated median or percentile, label it an estimate and avoid displaying spurious precision beyond the diagram.
The scale deserves attention. Volume 0099 on graph scales teaches how to read a truncated origin and equal or unequal tick intervals. Here the cumulative interpretation depends on reading both axes accurately before any subtraction.
A useful independent check is that the curve’s endpoint matches the total number of observations. If the plot ends at cumulative frequency 37 while the table contains 40 observations, something may have been plotted or read incorrectly. Fix that before locating quartiles.
6. How to find the median in grouped data
The median represents the middle of an ordered dataset. For 40 individual values, the central positions are the 20th and 21st observations. In a cumulative graph of grouped continuous data, learners commonly use half of the total cumulative frequency, 40/2=20, as the level at which to read an estimated median from the curve.
On our example, cumulative frequency reaches 13 by the boundary 40 and 28 by boundary 60. The middle lies in the interval above 40 and up to 60. That conclusion is secure from the grouped counts even before making a numerical estimate of the median within that class.
If a supplied cumulative curve is drawn in a way that supports reading a value at cumulative frequency 20, draw a horizontal line from 20 to the curve, then a vertical line down to the measurement axis. The horizontal coordinate is an estimated median. Do not mistake 20 for the median value itself; 20 is the cumulative count level being used to locate it.
For an illustrative straight-line interpolation between (40,13) and (60,28), the estimated value at CF 20 would be 40+(20−13)/(28−13)×20≈49.3. This result depends on an interpolation assumption within the class; the exact median of the underlying unseen observations is not determined by the grouped table alone.
That caveat is important. The graph or question may invite an approximate read rather than exact interpolation. Follow its conventions and specified accuracy. The purpose of this chapter is to understand why the answer is in a certain interval and how to read it, not to promote one invented calculation technique as a mandatory examination rule.
The median need not equal the mean. A distribution with some very large values may have a mean pulled upwards while the middle observation stays comparatively low. Cumulative-frequency diagrams help locate positions, while other summary statistics answer different questions.
7. Lower quartile, upper quartile and interquartile range
Quartiles divide an ordered dataset into four parts. The lower quartile is near the first quarter of the observations, while the upper quartile is near the third quarter. On a cumulative-frequency diagram, a common graphical approach uses levels around one quarter and three quarters of the total cumulative frequency. The exact conventions and data representation in the question matter.
For 40 observations, the lower-quartile level is around 10 and the upper around 30. Our cumulative total reaches 4 by boundary 20 and 13 by boundary 40, so the lower quartile lies within the interval above 20 and up to 40. The total reaches 28 by boundary 60 and 36 by boundary 80, so the upper quartile lies within the interval above 60 and up to 80.
These interval statements are supported by the grouped data. If the question supplies a graph, estimate the corresponding horizontal values using the appropriate levels. For illustration, a straight-line interpolation across the lower-quartile class gives 20+(10−4)/(13−4)×20≈33.3. Across the upper class it gives 60+(30−28)/(36−28)×20=65. These are grouped estimates, not exact quartiles of an unseen raw list.
The interquartile range, or IQR, is upper quartile minus lower quartile. Using the illustrative interpolations gives about 65−33.3=31.7 units. The calculation is meaningful as an estimated spread between the middle-half boundaries under the interpolation assumptions. Do not present it as an exact raw-data IQR unless the question or source supports that precision.
Why do we care? The IQR describes the spread of the middle portion of a distribution. It helps compare variation without letting a few extreme observations dominate the summary. It does not tell you everything about skewness or the number of observations at each exact score, but it offers one compact and useful measure.
A common mistake is to subtract the quartile positions themselves, such as 30−10=20, and call that the IQR. Those numbers are cumulative-frequency levels, not values on the measurement axis. You must read the corresponding lower and upper quartile measurement values before subtracting.
8. Percentiles: a position is not the measured score
A percentile describes a position in the ordered distribution. The 75th percentile is a value around which about three quarters of observations lie at or below it, subject to the calculation or graph-reading convention. The number 75 is not necessarily a score of 75.
Imagine 200 students taking an assessment. A cumulative frequency of 150 corresponds to three quarters of the group. If the graph shows CF 150 at a mark of approximately 68, then the 75th-percentile mark is about 68 on that graph. The percentile is 75; the measured value is about 68. Do not swap their axes.
Likewise, a score of 75 marks does not automatically mean the student is at the 75th percentile. If many students score higher or lower than 75, the percentile position may be quite different. The percentile must be found from the distribution, not copied from the score.
For a cumulative graph with total N, locate the specified fraction of N on the cumulative-frequency axis, then read the associated horizontal-axis value. If a question asks what proportion scored at most a certain value, move in the other direction: read the cumulative frequency at that value and divide by N.
Consider our 40-observation example. At most 60 accounts for 28 of 40 observations, or 70%. That says the 60 boundary lies around the 70th cumulative percentile under the grouping convention. It does not mean the average score was 70, nor that the 70th observation has score 60. There are only 40 observations in this example.
A reliable response names both axes: “At x=60, CF=28; proportion at or below 60 is 28/40=70%.” Small labels prevent a position from becoming mistaken for a measurement.
9. Comparing two distributions of different sizes
A raw cumulative count can mislead when two groups contain different numbers of observations. Suppose Class A has 40 learners, with 28 scoring at or below 60. Class B has 30 learners, with 24 scoring at or below 60. Which group has the larger proportion at or below 60?
For Class A the proportion is 28/40=70%. For Class B it is 24/30=80%. Class A has a larger cumulative count of 28, yet a smaller proportion. Comparing the numbers 28 and 24 without considering total size answers a different question from comparing percentages.
The difference matters when drawing conclusions about performance. If lower scores are less desirable in the stated context, Class B has a higher proportion at or below the 60 boundary. But the cutoff alone does not describe each class’s entire distribution. To compare their medians, quartiles or spread, look at the relevant parts of the cumulative curves or data.
Suppose the classes had identical percentile distributions but one was twice as large. Its cumulative counts at a given score would also be twice as large. It would be wrong to infer that students in the larger class performed better or worse merely from the taller cumulative-frequency axis.
To make a fair comparison, either interpret each group using its own total or convert cumulative counts to cumulative percentages when that is appropriate. Keep the populations and measurement scales comparable. A change of sample size does not alter the meaning of the variable on the horizontal axis.
This extends the argument in Volume 0079 about weighted comparisons. The question is not which number looks larger, but what is being counted and against which total.
10. Common errors and their direct repairs
The first error is treating a cumulative total as a class frequency. If CF increases from 13 to 28, the frequency in that class is 15, not 28. The repair is to subtract the lower cumulative total from the upper one.
The second is reading the wrong axis. To find a median from a cumulative curve, locate the relevant cumulative-frequency level and then read the measurement value. A frequency position is not itself a score or time.
The third is failing to check the last cumulative total. If class frequencies sum to 40 but the table ends at 36, you have either missed a class or made an addition error. The repair is to recompute the cumulative sums from the first interval.
The fourth is ignoring boundary conventions. A value exactly equal to a boundary must belong to the defined interval. The repair is to copy the interval notation or wording before subtracting.
The fifth is assuming every value within a grouped interval equals its midpoint. Midpoints are useful approximations for some grouped calculations, not direct evidence of each underlying value. A grouped quartile estimate cannot have more certainty than its source data.
The sixth is comparing cumulative counts from different sample sizes without normalising when the question asks for proportions. The repair is to use each group’s total as its denominator.
The seventh is interpreting a rising cumulative-frequency graph as proof that individual scores increase with time. The graph typically accumulates counts across increasing score or measurement boundaries. Its horizontal axis may not represent time at all.
Every error has a different cause. “I misread the graph” is too vague to guide practice. Identify whether you confused class and cumulative counts, positions and values, or populations and proportions.
11. Original practice questions, from basic to advanced
A. An ordinary-frequency table has class frequencies 3, 7, 10 and 5. Write the cumulative frequencies and state the total number of observations.
B. Cumulative frequencies are 5 by boundary 10, 14 by boundary 20 and 23 by boundary 30. Find the number of observations above 10 and at most 20.
C. A group contains 60 observations. The cumulative frequency at a boundary is 39. How many observations lie above that boundary under the stated at-most convention?
D. A cumulative graph for 50 observations reaches CF 25 at an approximate measurement value of 42. What does 42 represent in a typical graphical median-reading exercise? Why is 25 not the median measured value?
E. A cumulative curve for 80 observations reaches CF 20 near x=32 and CF 60 near x=72. Estimate the lower quartile, upper quartile and IQR.
F. Class A has 35 of 50 observations at most 60. Class B has 18 of 30 at most 60. Which has the greater proportion at or below 60?
G. A frequency table of 40 observations contains classes with frequencies 4, 9, 15, 8 and 4. Find the cumulative total after the third class and the frequency of the fourth class from the cumulative totals.
H. A student claims an upper quartile equals 75 because it is the 75th percentile. Explain the mistake.
I. A grouped table states that 12 observations are at most 40 and 26 are at most 60. What does 26−12 calculate? What wording would make the interval unambiguous?
J. A hand-drawn cumulative graph gives an estimated median of 48.7, but its small scale permits reading only to about the nearest unit. How should the student report precision responsibly?
K. Two classes have different sizes but similar cumulative percentages at every plotted score. Can you conclude from that alone that the larger class has a better distribution?
L. A graph has score on the horizontal axis and cumulative frequency on the vertical axis. A student says the line’s upward slope shows students improved over the term. What is the error?
M. A group of 40 has CF 13 at x=40, CF 28 at x=60 and CF 36 at x=80. Find the counts in the intervals above 40 to 60 and above 60 to 80. Which interval contains more observations?
N. A task asks for the number of students scoring at least 60, but the table gives the number scoring at most 60. Is a subtraction from the total enough to obtain the exact answer if students may score exactly 60? Explain.
Do not rush to a calculator. First mark the relevant boundaries and whether the question asks for a count, value, rank, percentage or estimate. Most mistakes are mistakes about meaning before they are mistakes about subtraction.
12. Answer clinic with explicit reasoning
A. Cumulative frequencies are 3, 10, 20 and 25. The total number of observations is 25, matching 3+7+10+5.
B. The interval count is 14−5=9. Under the at-most convention, this is the number above 10 and at most 20.
C. Observations above the boundary equal 60−39=21, because 39 have already been counted at or below it.
D. The horizontal value about 42 is the estimated median measurement. The cumulative level 25 locates the middle of 50 observations on the vertical axis.
E. Estimated lower quartile is 32, upper quartile 72 and IQR is 72−32=40 units. These are graph-based estimates rather than exact raw-data values.
F. Class A has 35/50=70%. Class B has 18/30=60%. Class A has the greater proportion at or below 60, even though its raw cumulative count is also larger.
G. The total after the third class is 4+9+15=28. The next cumulative total is 36, so the fourth-class frequency is 36−28=8.
H. The 75th percentile describes the rank position in a distribution. The upper-quartile measurement depends on the data and need not equal the number 75.
I. The difference 26−12=14 counts values above 40 and at most 60 under that convention. Writing 40 < x ≤ 60 preserves both boundaries.
J. Report a sensible approximate reading such as about 49, or to the resolution genuinely supported by the graph and question. Do not claim the tenths digit is exact if the drawing cannot support it.
K. No. Similar cumulative percentages suggest similar relative distributions at the plotted boundaries, not better performance by the larger class merely because it contains more people.
L. The graph accumulates counts across the measurement or score axis. If the horizontal axis is score, it is not a time graph and cannot by itself establish improvement over a term.
M. From 40 to 60, count 28−13=15. From 60 to 80, count 36−28=8. The first interval contains seven more observations.
N. No, not necessarily. Total minus the number at most 60 yields those strictly greater than 60. “At least 60” also includes scores equal to 60. You need the count at exactly 60 or a suitable alternative cumulative boundary to answer precisely.
Mark your errors by what they reveal. A wrong answer to Question N is a boundary-language error. A wrong answer to Question E is likely a quartile-axis error. The next practice should match the actual gap.
13. A worked study routine for grouped distributions
Begin with a small table of five classes and simple integer frequencies. Confirm the total. Build cumulative totals and reconstruct ordinary frequencies by subtraction. Repeat this forward-and-back process until neither direction is mysterious. This is the foundation that later quartile questions rely on.
Next draw a rough cumulative curve using the class boundaries and cumulative totals. Label both axes with the correct variable, unit and cumulative frequency. A rough sketch can teach the relationships without requiring a perfect artistic curve. Your first aim is accurate meaning and position, not visual decoration.
Then practise locating N/2, N/4 and 3N/4 levels. For N=40 these are 20, 10 and 30. Read the corresponding horizontal values from a supplied suitable curve, and identify when you are making a grouped estimate. Do not present an interpolated result as if you possess the full raw dataset.
For comparison tasks, use two distributions with different totals. Find relevant cumulative percentages, not only counts. Explain which group has a higher proportion below a defined boundary and how that differs from comparing medians or quartiles. Make the question’s meaning explicit.
Finally answer one unfamiliar word problem that embeds cumulative frequency in a realistic setting: time taken to complete a task, measurement ranges, travel distances or scores. Translate the story into the correct data interpretation, show the calculation and return a sentence with units.
This process is deliberate, not slow for its own sake. Once the core relationship is secure, many short questions become easier because they no longer require guessing which number belongs in the answer.
14. A seven-day progression with independent checks
Day 1: Convert three ordinary-frequency tables to cumulative form. Check that each final cumulative total equals the sum of frequencies.
Day 2: Recover class frequencies from cumulative totals. Use at-most and above-boundary wording to identify the correct subtraction.
Day 3: Read cumulative curves with clearly labelled axes. Describe the meaning of five plotted points in complete sentences before calculating anything else.
Day 4: Locate the median and quartile levels. Explain why cumulative positions and measurement values belong to different axes. Practise using an appropriate approximate graph read.
Day 5: Compare different-sized groups using cumulative percentages. Verify that the denominator is the correct group total and do not infer unrelated causal conclusions.
Day 6: Diagnose misleading graphs or ambiguous interval boundaries. Return to Volume 0099 for scale problems and Volume 0101’s integrated evidence guide when the stated data cannot answer a broader claim.
Day 7: Attempt Questions A–N independently and review the explanation for each error. Try a second set several days later with changed class boundaries and totals. If you can still select the correct cumulative operation, the learning has transferred.
For parents and teachers, a useful question is: “Does that point count the people in this interval or everyone counted so far?” If the learner can answer and demonstrate by subtraction, they understand the basis of the topic.
15. The final link: the graph is a count of a set
The central skill is not drawing a beautiful curve. It is understanding what each value includes. A cumulative frequency at a boundary counts all observations in the accumulating set. Moving from one boundary to another adds a new interval. Subtracting the old count from the new count reveals the interval. Locating a percentile means selecting a cumulative position and then reading a measurement value.
Keep counts, values, rankings, fractions and graph appearances separate. That discipline helps with statistical reasoning far beyond one style of school question. It also protects learners from persuasive but misleading charts in everyday life.
Continue through the Mathematics Hub and G3 SEC Learner’s Guide. If the difficulty begins with fractions, percentages or drawing axes, repair those foundations first using the appropriate existing route rather than restarting the entire syllabus. If the issue is how far a conclusion may travel beyond its evidence, Volume 0093 on sequence and cause provides a useful companion.
Official references: SEAB 2027 G3 syllabus directory and 2027 G3 Mathematics K310 syllabus. These explanations, numbers and questions are original educational practice, not reproduced SEC questions.
