Quick answer. Compare G3 SEC Mathematics data by choosing the right measure: mean and median describe centre; range, IQR and standard deviation describe spread; cumulative frequency shows how many observations fall below stated thresholds. Correct interpretation depends on units, boundaries and comparable samples.
The quick answer: a typical result is not the whole result
When two groups complete the same kind of science-laboratory task, the first question people ask is often, “Which group was faster?” A single average appears to settle the matter. But it can hide very different patterns. One group may finish most tasks at about the same pace while another finishes a few extraordinarily quickly and a few very slowly. Those distributions demand different interpretations, especially when the work concerns careful measurement or safe fault diagnosis rather than a simple race.
This guide studies time records from an invented classroom circuit-fault exercise. It moves beyond the current–voltage gradients in Volume 0614 into statistical comparison. You will learn to identify a useful average, measure spread, draw a correct cumulative-frequency description, inspect a box-and-whisker plot, and decide what the numbers do not establish.
The strongest examination answer is neither a pile of correct calculations nor an unsupported verdict such as “Team A is better.” It states the measure, compares like with like, explains what a larger or smaller value means, and acknowledges an important limitation. That habit matters in Science, Mathematics and English. The same caution is needed when diagnosing faults: one current reading does not identify every possible cause, and one median does not describe every aspect of a group’s performance.
All times, groups and examples here are fictional learning data. They do not describe actual students or published school results. The objective is a transferable mathematics method, not a league table of people.
What the 2027 G3 Mathematics examination tests
For the 2027 Singapore-Cambridge Secondary Education Certificate, G3 Mathematics is syllabus K310. The earlier 4052 code is supplied by SEAB as a reference for 2026 and previous examinations, not as the current 2027 code. The official K310 syllabus has two papers. Each lasts two hours and 15 minutes, carries 90 marks and contributes 50% of the examination. Paper 1 contains approximately 26 short-answer questions. Paper 2 contains 9 to 10 questions of varying lengths and marks, and its final question focuses on applying Mathematics to a real-world situation.
The syllabus explicitly includes collecting, classifying and presenting data; tables; multiple graph types; cumulative-frequency diagrams; box-and-whisker plots; mean, median and mode; quartiles; range; interquartile range; standard deviation; and comparing datasets. The real-world-question provision makes this chapter’s focus appropriate: students must do calculations and explain their relevance to the situation in which the data were gathered.
The official notes require essential working; omitting it can cost marks. Non-exact numerical results are normally given to three significant figures, while angles in degrees are usually given to one decimal place, unless a question specifies another accuracy. Keep exact intermediate values when sensible and round at the requested stage. If the task supplies its own rounding instruction, follow that instruction. This is a syllabus-based guide and not an official examination paper; it does not claim to reproduce a particular future question.
If you are returning from PSLE Mathematics, the main change is that a numerical answer increasingly needs interpretation. Earlier work with averages and percentages gives you an entry point. G3 statistical reasoning asks how a whole set behaves, whether a comparison is fair, and what could make an apparently convincing figure misleading.
Collect comparable data before you calculate
Imagine two teams doing eight supervised paper-and-pencil simulations of circuit-fault diagnosis. We record how many minutes each team takes to identify the best-supported fault hypothesis in each exercise. Team A records 2, 3, 3, 4, 4, 5, 6 and 7 minutes. Team B records 1, 1, 2, 4, 5, 6, 9 and 12 minutes. The times are already placed in ascending order to make the calculations clearer.
A data table is not automatically a fair experiment. Did both teams receive questions of comparable difficulty? Did they use the same definition of “completed”? Did a teacher count a partly correct explanation as a finish? Did the clock include reading time? Were there distractions or differently equipped workspaces? A sound mathematical analysis begins by recognising the variable and its measurement conditions. In this scenario the variable is completion time, measured in minutes, for a defined type of task.
The data have exactly eight observations for each team, but equal sample size does not establish comparable task allocation. If Team B received more complicated networks, comparing raw completion time alone may be unfair. The calculations remain mathematically correct; the interpretation must remain conditional. In an examination, the given assumptions may state that tasks were equivalent. You should accept stated conditions while avoiding claims beyond them.
Before doing any arithmetic, scan the data. Team A’s values cluster between three and six minutes. Team B includes two very short times and a particularly long result at twelve minutes. That visual difference suggests why mean, median and spread may not all tell the same story. It is a preview, not a conclusion.
Find the mean without letting one long time vanish
The arithmetic mean is total divided by number of observations. For Team A, the total is 2 + 3 + 3 + 4 + 4 + 5 + 6 + 7 = 34 minutes, so the mean is 34/8 = 4.25 minutes. For Team B, the total is 1 + 1 + 2 + 4 + 5 + 6 + 9 + 12 = 40 minutes, so the mean is 40/8 = 5.00 minutes. By the mean, Team A’s recorded time is lower.
The mean is sensitive to every data value, including unusually large observations. Team B’s twelve-minute task adds substantially to its mean. This does not make the mean wrong. It makes the mean a particular summary: total time per observation. That can matter if the goal is to estimate the total time a team will need across many similar tasks. It may be less useful as a description of what a typical individual task felt like.
A common mistake is to compare a mean calculated from one time unit with a median reported in another. Keep units consistent. A second mistake is to say the mean is necessarily a value observed in the data. Neither team has a 4.25-minute observation, and that is not a problem. An average can fall between the recorded values.
In your written conclusion, name the measure: “Team A had the lower mean completion time, 4.25 minutes compared with 5.00 minutes.” This is more precise than “Team A was faster” because the latter may suggest that Team A completed every individual task faster. It did not. Team B has one-minute observations, shorter than any Team A time.
Find the median as the middle position
For an even number of ordered observations, the median is the average of the two middle values. Both teams have eight results, so use the fourth and fifth observations. Team A’s fourth and fifth values are 4 and 4, giving a median of 4 minutes. Team B’s fourth and fifth values are 4 and 5, giving a median of 4.5 minutes.
The median is less sensitive than the mean to how far an extreme observation lies from the centre. If Team B’s twelve-minute time were sixteen minutes instead, its median would remain 4.5 minutes while its mean would rise. The median therefore helps describe a central position when long tails or large values exist. It does not erase difficult tasks; it answers a different question.
Students sometimes say that half the observations must be below the median and half above it. That wording becomes awkward when values are tied. A safer description is that the median marks the middle position of an ordered set, dividing its ranked observations into two halves. Some values may equal the median. The median of the eight results is calculated by the standard middle-pair method, not by guessing the most frequent number.
The comparison now contains two related but distinct statements. Team A has a median of 4 minutes versus Team B’s 4.5 minutes; Team A also has the lower mean. The medians are relatively close, while the means differ by 0.75 minutes. Neither comparison shows how consistent the teams were, so we turn to spread. A good answer knows when its current statistic has finished its job.
Quartiles: look inside the lower and upper halves
Quartiles describe the positions of the data after ordering. For these even-sized examples, we use the common median-of-halves convention taught in many school exercises: find the median of the lower four observations for Q1 and the median of the upper four for Q3. Always use the convention specified by the examination question or your teaching materials when a method is given; different statistical software may define sample quartiles differently.
For Team A, the lower half is 2, 3, 3, 4, so Q1 = (3 + 3)/2 = 3 minutes. The upper half is 4, 5, 6, 7, so Q3 = (5 + 6)/2 = 5.5 minutes. For Team B, the lower half is 1, 1, 2, 4, so Q1 = (1 + 2)/2 = 1.5 minutes. The upper half is 5, 6, 9, 12, so Q3 = (6 + 9)/2 = 7.5 minutes.
Do not treat a quartile as the arithmetic mean of one-quarter of the dataset. Q1 and Q3 are positions in the ordered distribution. Their usefulness lies in describing how widely the central part of the dataset is spread. We need them both to calculate the interquartile range.
A student can check the values against the overall median. Q1 should not exceed the median, and Q3 should not fall below it in this ordered context. The inequalities are not a complete check, but they catch impossible quartiles caused by choosing the wrong half. When there are many values, keep their positions visible rather than writing a sequence of unexplained figures. Examination working should let the marker follow the selection of the quartile positions.
Interquartile range and range answer different questions
The interquartile range, or IQR, is Q3 − Q1. Team A has IQR = 5.5 − 3 = 2.5 minutes. Team B has IQR = 7.5 − 1.5 = 6.0 minutes. The central half of Team A’s recorded times is therefore more tightly grouped by this measure. That is a meaningful statement about consistency in this sample.
The range is maximum minus minimum. Team A has range = 7 − 2 = 5 minutes. Team B has range = 12 − 1 = 11 minutes. The range confirms that Team B’s total span of observed completion times is much wider. But because the range depends only on two extreme observations, it can change dramatically when one new extreme is added.
Neither IQR nor range is a measure of how fast a team is on average. A group can have a small IQR and a high median if all its times are clustered around a slow performance. A group can have a large IQR but a low median if half its tasks are completed rapidly and the others take much longer. The correct comparison names the dimension: central location, middle-half spread, or full span.
The sentence “Team A was more consistent because its IQR was 2.5 minutes, compared with 6.0 minutes for Team B” is defensible for these observations. “Team A will always be more reliable” is not. That broader claim would require a suitable definition of reliability, an adequate sample and reasonable control of task difficulty. The mathematical measure is strong evidence of a distributional difference in the sample, not a guarantee about every future outcome.
Build a box-and-whisker picture from five numbers
A box-and-whisker plot is a compact representation built from the minimum, lower quartile, median, upper quartile and maximum. For Team A these five numbers are 2, 3, 4, 5.5 and 7. For Team B they are 1, 1.5, 4.5, 7.5 and 12. On one shared numerical scale, draw a box from Q1 to Q3, place a line at the median, and extend whiskers towards the minimum and maximum using the plotting convention specified in the question.
The width of the box represents the IQR. Team B’s box is longer, showing greater spread in its central half. Its right whisker is also longer, matching the presence of long completion times. A student does not need to invent a precise average from the box plot, because the box plot does not display the sum of all data values. It displays selected positions.
Pay attention to the axis. If two separate plots use different scales or starting points, visual width cannot be compared honestly by eye. The correct comparison requires the same scale or reading the numerical five-number summaries. In a paper, a carefully labelled axis and clear marks are part of mathematical communication, not decoration.
The plot also gives useful information about asymmetry. Team B has a long upper tail, but the five-number summary alone is not enough to identify an underlying probability distribution or explain the causes of the long times. The correct verbal result is modest: the observed upper half of Team B’s data is more spread out. The explanation for that pattern must come from the task design or additional evidence.
Grouped tables: preserve the boundaries
Suppose we group times into equal-width intervals: 0 ≤ t < 5 minutes, 5 ≤ t < 10 minutes, and 10 ≤ t < 15 minutes. For Team A, the counts are 5, 3 and 0. For Team B, they are 4, 3 and 1. Check that each team's frequencies add to eight. The intervals are half-open: a time of exactly 5 minutes belongs to the second class, not the first.
This boundary issue is easy to overlook in examinations. If a question labels intervals as 0–4, 5–9 and 10–14 for integer-minute records, no value can belong to two classes. But measurements may be decimals, so boundaries such as 0 ≤ t < 5 are clearer. The mathematical statement about included endpoints matters more than the typographic dash.
Grouped data sacrifice information. From the frequency table alone, we know how many observations fall into each class, but not precisely which values they are. We cannot reproduce the exact mean of the ungrouped dataset without further assumptions. If asked to estimate a grouped mean, use class midpoints and identify the result as an estimate. Do not silently treat it as the exact mean we computed from the full records.
A histogram with equal-width classes can represent the frequencies of these intervals. Label the horizontal axis as time in minutes and the vertical axis as frequency, using the appropriate instructions for the graph. No bar should overlap the neighbouring class incorrectly. The visual goal is to show the shape of the distribution, not to make one team appear better through a convenient axis scale.
Cumulative frequency: answer “how many by this time?”
Cumulative frequency adds successive class frequencies. Using the classes above, Team A has 5 observations below 5 minutes, 8 below 10 minutes and 8 below 15 minutes. Team B has 4 below 5, 7 below 10 and 8 below 15. At the last upper boundary, both totals reach the sample size of eight. A cumulative total that exceeds eight indicates a counting error.
A cumulative-frequency graph is built from upper class boundaries and cumulative totals, together with a suitable starting point. For these half-open intervals, points include (5, 5), (10, 8), (15, 8) for Team A, and (5, 4), (10, 7), (15, 8) for Team B. If the data set contains only eight observations, this exercise is an illustration of the method; larger grouped datasets make smooth curves and percentile estimates more useful.
Do not confuse “less than five minutes” with “five minutes or less.” Our first class excludes exactly five. If a report says “five or fewer,” you must count the five-minute observations too. Team A then has six results at most five minutes, while Team B has five. This is not a contradiction; it is a change in the question’s boundary.
Cumulative frequency helps answer positional questions: what proportion of results fell below a threshold, at roughly what time a given percentile is reached, or how two groups’ accumulation of results compares. It does not directly explain why one task was long. Nor does the cumulative curve show which person or particular task generated the extreme value. Its strength is description of a distribution through ordered counts.
Percentages require the right denominator
Using the eight-task logs, five of Team A’s tasks took less than five minutes. That is 5/8 × 100% = 62.5%. Four of Team B’s tasks took less than five minutes, which is 4/8 × 100% = 50%. The difference is 12.5 percentage points. Do not say it is automatically a 12.5% relative improvement. Percentage points compare percentages by subtraction; relative percentage change requires choosing a reference percentage and dividing by it.
Now consider a different question: among all tasks that took at least five minutes, which team’s recorded cases were more numerous? Team A has three such cases and Team B has four. If the dataset combines both teams, there are seven tasks in that time category, of which Team B accounts for four. The relevant proportion is 4/7, about 57.1%. It is not 4/8, because the question has changed the denominator to all tasks meeting the condition.
A reliable method is to write the denominator in words before performing a calculation. “Out of all Team B tasks” leads to eight. “Out of all long tasks across both teams” leads to seven. “Out of all tasks in the complete dataset” leads to sixteen. Numbers alone do not tell you which base is correct.
If a task asks about a proportion and does not define the relevant total clearly, examine the wording. Count the set named by “among,” “of those” or “given that.” These phrases are as important in data interpretation as they are in English comprehension. Mathematics is precise language joined to quantities. A denominator error is often a reading error before it becomes an arithmetic error.
What standard deviation adds—and what it does not
Standard deviation measures spread around the mean using every observation. It is especially useful when a question asks students to compare central tendency and variability together. The 2027 K310 syllabus includes the calculation of standard deviation for grouped and ungrouped data, together with using mean and standard deviation to compare sets.
For an ungrouped population-style dataset, one convenient expression is the square root of [sum of x² divided by n, minus (mean)²]. For Team A, sum of squares is 4 + 9 + 9 + 16 + 16 + 25 + 36 + 49 = 164. Divide by eight to get 20.5. Subtract 4.25² = 18.0625, giving variance 2.4375. The standard deviation is √2.4375, approximately 1.56 minutes to three significant figures.
For Team B, the squared values sum to 1 + 1 + 4 + 16 + 25 + 36 + 81 + 144 = 308. Divide by eight to get 38.5. Subtract 5² = 25, giving variance 13.5. The standard deviation is √13.5, approximately 3.67 minutes. As with IQR, Team B has the greater spread, but the two measures use the data differently.
Check which standard-deviation definition the question expects. Statistical software can use a sample estimator with n − 1 in the variance formula; the G3 school syllabus provides the population-style calculation formula for the data set in the question. Do not mix conventions silently. A standard deviation has the same unit as the original values; here the unit is minutes, not minutes squared. Variance has squared units. That distinction can help catch an answer that stopped one operation too early.
A worked real-world decision without pretending it is certain
An exhibition organiser needs a team to demonstrate circuit fault reasoning within a short session. Team A has the lower mean, lower median, lower IQR and lower standard deviation in this eight-task sample. On the evidence given, it is reasonable to prefer Team A if the priority is a more predictable completion time, assuming both groups faced comparable tasks and accuracy was acceptable.
However, time is not the only quality measure. A group might be fast because it guessed and received credit for incomplete explanations. A safer study would record whether each diagnosis was correct and whether the evidence chain was adequate. The organiser should avoid treating speed as a proxy for safety, comprehension or teaching ability.
Suppose the decision rule says “select a group only if its median time is no more than five minutes and it has no incorrect safety claims.” Both groups meet the median-time condition, but the supplied time records do not tell us whether either group made incorrect safety claims. The choice cannot be completed from the times alone. A good mathematical answer says which condition can be evaluated and identifies the missing information.
In long real-world questions, students sometimes use every number in the statement because they think every number must enter the same calculation. That is not always the purpose. Some data identify a constraint; some provide context; others test whether a conclusion is justifiable. Translate the decision rule into separate tests, calculate the relevant measures, and apply the rule faithfully. A mathematically correct statistic used to answer the wrong decision question is not a complete solution.
Six examination-style drills with reasoning
Drill 1. Find the median of 3, 5, 7, 7, 9. The ordered middle observation is 7, so the median is 7. It is already an odd-sized set; there is no need to average two values.
Drill 2. A set has minimum 2, Q1 = 4, median 6, Q3 = 10 and maximum 12. Find the IQR and range. IQR = 10 − 4 = 6; range = 12 − 2 = 10. Both have the same unit as the original data. The five numbers are sufficient for these two calculations, though not generally for the mean.
Drill 3. A table has frequencies 4, 7, 5 and 2. Find the cumulative frequencies. They are 4, 11, 16 and 18. The final total shows eighteen observations. If you were given an original sample of sixteen, the mismatch would demand checking the table before drawing a graph.
Drill 4. One team has a median of six minutes and IQR of one minute; another has median four minutes and IQR of five minutes. Which was more consistent in its central half? The first, by IQR. Which had the shorter typical time by median? The second. One team need not win on both dimensions.
Drill 5. Out of 25 tasks, eight took more than seven minutes. Find the percentage. It is 8/25 × 100% = 32%. If the next question asks the percentage among those needing a second check, the denominator must be changed to the number needing that check; 25 is not automatically correct.
Drill 6. A question reports that a dataset’s mean is higher than its median. Must every value above the median be an outlier? No. A longer upper tail or some large values can pull the mean upward; identifying outliers needs an appropriate criterion and sufficient information. The two measures alone do not prove a particular observation is erroneous.
For each drill, practise a two-part answer: calculation, then one sentence of meaning. Do not write a page where a straightforward numerical question needs only a number and units, but do not omit explanation when the command word asks for comparison or interpretation.
A corrected answer is more valuable than an extra page of questions
Students often practise calculating quartiles successfully but lose marks when explaining them. The correction is not necessarily another twenty quartile questions. It is a short comparison in which the student must state what the quartiles reveal and what they leave unknown. A teacher can help by asking the student to identify which sentence would still be true if one extreme value changed.
Use a four-column error log: question demand, original mathematical move, corrected move, and reason. For example, a student wrote “Team A is always faster” after finding a lower mean. The corrected move is “Team A has the lower sample mean.” The reason is that a mean does not compare every task individually and does not warrant an absolute future claim.
For a histogram mistake, the original move might place a value of exactly five minutes in the class below five. The corrected move is to read the boundary notation before tallying. For a cumulative-frequency mistake, the original move might copy class frequencies without adding them. The corrected move is to keep a running total and verify the last count equals the sample size.
For a standard-deviation mistake, the student may stop at the variance and label it minutes. The repair is to take the square root and attach the original time unit to the resulting standard deviation. Work backwards from the error’s cause, not forwards from the answer key. That is how a single corrected question starts to improve future work.
A weekly rhythm is enough for many learners: one day of exact raw-data calculations, one day of graph-reading, one day of short written comparisons, one day of unfamiliar contextual questions, and a final day of independent retry. Pace matters less than whether the student can perform the repaired step without a model answer in front of them.
From PSLE arithmetic to G3 statistics and beyond
A student arriving from PSLE knows that an average can summarise several numbers. Secondary statistics adds a deeper question: in what sense does that average represent the whole group? The mean reflects every value; the median marks the centre; the IQR shows central spread; the standard deviation describes spread around the mean; cumulative frequency shows how counts accumulate across ordered thresholds. None is interchangeable with the others.
Read the PSLE Mathematics word-problem guide for the earlier task of translating prose into mathematical relationships. The PSLE-to-secondary learning plan helps families identify the transition to more formal explanation. For ongoing subject work, see the Secondary Mathematics statistics and probability guide and the graphs and functions guide.
This chapter follows Volume 0617, which distinguishes circuit faults from meter evidence. It does not replace that Physics explanation; it addresses a different mathematical intent: comparing logs of repeated work. The next English chapter asks how to present the results to a reader without converting a limited sample into an exaggerated claim.
Return to the G3 SEC EMS learner-guide hub when you need to repair an earlier skill. Strong learning routes make it possible to travel backwards to a weak prerequisite, then forwards again to a more demanding question.
Official sources and a simple final test
For the current qualification and subject codes, consult the SEAB SEC introduction and the 2027 G3 school-candidate syllabus page, last updated 26 October 2026. For the Paper 1 and Paper 2 formats, statistical topics and examination conventions, read the official K310 Mathematics syllabus. This chapter is a learning guide with original examples, not a specimen paper or a source of guaranteed future questions.
Before finalising a statistics answer, check five things: the values belong to comparable groups; the calculation uses the correct observations; any percentage has the intended denominator; the graphical scale and class boundaries are correctly labelled; and the written conclusion refers only to what the measure can show. A smaller number is not automatically better, and a correct number is not automatically the whole answer. Mathematics becomes useful when the conclusion respects what was measured.
