Did you know? Two groups can have the same median yet be very different in how spread out their results are. A single unusually slow observation can move a mean and stretch a graph, even though most of the group did not change. This is why a Mathematics reader should inspect both centre and spread before deciding what data mean.
In this fictional library study, twelve people are timed while locating a specified resource. We will calculate mean, median, range, quartiles, interquartile range and standard deviation. We will then decide whether a stem-and-leaf diagram, box plot, histogram or honest numerical table best communicates the result.
The study is an invented Secondary 4 G3 Mathematics learning exercise for Bukit Timah families. It is not a real survey of a library, actual people or the performance of any school. The 2027 SEC G3 Mathematics K310 syllabus explicitly includes statistical summaries, diagrams, comparisons and explaining misleading representations.
First read the measurement and the sample size
Our fictional measure is minutes needed to locate one specified library resource. Smaller recorded times describe faster completion under this model, but they do not automatically reveal why one person took longer.
Twelve observations are supplied. The group is not a random representative sample of any city or school. It is a teaching data set designed to make calculations auditable. Do not claim it measures a real library or predicts outcomes for all students.
The unit is minutes, not number of visits, number of books or a subjective satisfaction score. A statistical answer should use the correct variable and unit throughout.
The twelve observations in ascending order
Group A’s invented times are 6, 7, 8, 8, 9, 10, 10, 11, 12, 13, 14 and 24 minutes. They are already ordered from smallest to largest.
The repeated 8 and repeated 10 represent separate recorded observations. Do not collapse duplicates into a single number, because doing so would change the sample size and the mean.
The last value, 24 minutes, is much larger than most others. It deserves attention when we study the range and mean, but the data alone do not tell us whether it arose from a difficult search task, a delay or any other cause.
A count check before any statistical formula
There are twelve values in Group A. A quick manual count can label them positions 1 through 12; the sixth and seventh entries will control the median.
Write the sample size n = 12 in the margin. It is different from the sum of observed times, which will be much larger than twelve. The mean requires total minutes divided by twelve observations.
Did you know? Counting an observation twice or accidentally omitting one entry can make every subsequent statistic inaccurate even when the formula used is correct.
Mean of Group A: account for every value
Add the observations: 6+7+8+8+9+10+10+11+12+13+14+24 = 132 minutes. There are twelve results, so the mean is 132/12 = 11 minutes.
This mean describes the arithmetic average time of the twelve fictional observations. It does not mean every person took eleven minutes, and it does not guarantee that a new person would take eleven minutes.
A useful check is 12 × 11 = 132, matching the total. Write the unit after the final mean; reporting merely ’11’ leaves the interpretation incomplete.
Median of Group A: the centre of the ordered list
With twelve observations, the median is the average of positions six and seven. Both values are 10, so median = (10+10)/2 = 10 minutes.
The median is one minute lower than the mean. One reason the mean is higher is that the 24-minute result contributes a comparatively large value to the sum.
Do not simply choose position six without checking whether there is an even number of observations. If the two central values had differed, their average would be required under this ordinary median definition.
Mode: a data set may have more than one
The time eight minutes occurs twice, and ten minutes also occurs twice. No other value appears as often, so this Group A data set has two modes: 8 and 10 minutes.
A mode is the most frequently occurring observed value, not necessarily the middle or average. A data set may have one mode, several equally common modes or no uniquely most frequent value.
The duplicated entries should remain visible in a stem-and-leaf or dot diagram. Erasing one because it ‘looks repeated’ would erase a real observation in this fictional sample.
Range: a simple measure of overall spread
The minimum is 6 minutes, and the maximum is 24 minutes. The range is therefore 24−6 = 18 minutes.
This is a valid, quick calculation, but it is strongly affected by the highest recorded value. Range alone does not say how the ten or eleven central observations are arranged.
Did you know? Two data sets may have exactly the same range yet very different patterns inside it. A range is useful, but it is not a complete description of consistency.
Define a quartile convention before calculating
Different statistical systems and software can use different conventions for quartiles, particularly with odd samples or interpolation. For these classroom calculations, use the median-of-halves convention.
Since Group A has twelve values, split the ordered list into the lower six and upper six. Find the median of each half. This procedure defines Q1 and Q3 for this illustration.
If a national examination task explicitly gives quartile values or a particular method, follow that instruction. Clearly stated conventions are more mathematically honest than claiming every possible software package must give the same boundary.
Lower quartile of Group A
The lower six observations are 6, 7, 8, 8, 9, 10. Their middle values, the third and fourth, are both 8.
Therefore the lower quartile is Q1 = (8+8)/2 = 8 minutes. It represents the lower boundary of the central part of the ordered data under our stated convention.
A pupil who chooses the third entry of the complete twelve-number list without first locating the lower half may find the wrong quartile position.
Upper quartile and interquartile range
The upper six observations are 10, 11, 12, 13, 14, 24. Their two middle values are 12 and 13, giving Q3 = 12.5 minutes.
The interquartile range is Q3−Q1 = 12.5−8 = 4.5 minutes. It describes the width of the central half of this distribution under our quartile procedure.
Contrast the IQR of 4.5 with the total range of 18. The central observations are much more concentrated than the extreme endpoints suggest.
The five-number summary for a box-and-whisker plot
For Group A the five-number summary is minimum 6, Q1 8, median 10, Q3 12.5, maximum 24, all in minutes.
In a simple minimum-to-maximum school box plot, the box stretches from 8 to 12.5, a central median line appears at 10, and the whiskers extend to 6 and 24.
Some statistical software uses different whisker conventions and marks outliers separately. When comparing school diagrams, read the supplied plot rules and axes rather than assuming every display must use the identical convention.
A second fictional group with the same median
Group B’s twelve ordered times are 8, 9, 9, 9, 10, 10, 10, 10, 11, 11, 11, 12 minutes.
The sixth and seventh observations are both 10, so its median is also 10 minutes. Group A and Group B share exactly the same median even though their lists look different.
If a student concludes both groups have identical performance solely from their shared median, they have not examined the spread or the rest of the values.
Calculate Group B’s mean and range
The total of Group B’s twelve observations is 120 minutes: 8 + three 9s + four 10s + three 11s + 12. Therefore its mean is 120/12 = 10 minutes.
Its minimum is 8 and maximum is 12, so its range is 12−8 = 4 minutes. That is much smaller than Group A’s range of 18.
It is accurate to state that B’s observed times are less dispersed by range in this fictional sample. It is not yet justified to state why, or claim one teaching method caused it.
Calculate Group B’s quartiles and IQR
Under the same median-of-halves convention, the lower six Group B values are 8, 9, 9, 9, 10, 10. The middle pair are both 9, giving Q1 = 9 minutes.
The upper six are 10, 10, 11, 11, 11, 12. Their middle pair are both 11, giving Q3 = 11 minutes. The IQR is 11−9 = 2 minutes.
Group B has IQR 2 compared with Group A’s 4.5. Their medians are identical, so a good comparison names both the shared centre and the contrasting spread.
An accurate comparison sentence for an examination
Both groups have a median of ten minutes. Group B’s central observations are more tightly clustered because its IQR is 2 minutes, smaller than Group A’s 4.5 minutes.
Group B also has a smaller total range, 4 minutes versus 18 minutes. These are descriptive comparisons supported by the data.
Avoid saying Group B’s participants are inherently more capable or were taught better. The fictional sample does not establish those causal claims. Statistical communication means stating no more than the numbers support.
One important question: how far did each time lie from the average?
The Group A mean was 11 minutes. A person taking 6 minutes lies 5 minutes below it, while a person taking 24 minutes lies 13 minutes above it. This helps explain why the group has a larger spread than a cluster of times around 10 minutes. A single summary number cannot show all those distances.
To move from this observation to standard deviation, we use the squared differences from the mean. Squaring makes each contribution nonnegative and gives larger deviations more influence. The result summarises the spread of the whole stated collection of observations, rather than identifying whether any single student performed well.
Did you know? A large standard deviation does not mean a large average. The two statistics answer different questions. One tells us the centre of the numerical data, while the other describes variation around that centre.
Worked standard deviation of Group A
The twelve Group A values total 132, so their mean is 11. Their sum of squared values is 1700. Using the population standard-deviation form √[Σx²/n − (Σx/n)²], the variance is 1700/12 − 11².
That is 141.6667 − 121 = 20.6667 square minutes. Taking the square root gives standard deviation approximately 4.55 minutes, to three significant figures. These intermediate calculations are retained with additional precision before the final rounding.
Do not confuse the variance’s squared units with the standard deviation’s ordinary minute units. Also distinguish the population formula used for this complete illustrative twelve-observation collection from a different sample-standard-deviation calculation that software may offer.
Worked standard deviation of Group B
The Group B times total 120, giving mean 10 minutes. The sum of their squared values is 1214. The same population expression gives variance 1214/12 − 10² = 101.1667 − 100 = 1.1667 square minutes.
Taking the square root produces standard deviation approximately 1.08 minutes. Group A’s value was about 4.55 minutes, so B also has a smaller spread according to standard deviation, not only according to range and IQR.
This is a defensible comparison for the two fictional groups. It does not allow us to conclude that one group received better teaching, because neither sampling conditions nor causes were established in the exercise.
What the published SEC formula expects the student to understand
The 2027 G3 Mathematics K310 syllabus includes standard deviation and a formula based on frequency data, commonly written √[Σfx²/Σf − (Σfx/Σf)²]. Here f is frequency and x the observed value or class midpoint as appropriate. When each observation is written separately, each has frequency one.
For example, Group A’s value 8 occurs twice. In a frequency table that row contributes 2×8 = 16 to Σfx and 2×8² = 128 to Σfx². Removing the duplicate would change both sums and therefore the answer.
A student should be able to explain what the frequency does before using a calculator statistics function. The machine can evaluate the expression but cannot verify that the source table was interpreted correctly.
Build a stem-and-leaf diagram without losing repeated values
All the fictional times are whole minutes, so a stem-and-leaf diagram is a compact way to preserve them. For Group A, a key can say 1 | 2 means 12 minutes. The one-digit observations can be shown on stem 0 and two-digit observations on stems 1 and 2.
The ordered rows are 0 | 6 7 8 8 9, 1 | 0 0 1 2 3 4, and 2 | 4. Count the leaves: five plus six plus one gives twelve observations, which matches the original list.
This presentation makes the repeated 8 and 10 easy to see. Unlike grouped class intervals, it retains each value exactly, so the median and range can still be recovered without guessing observations from a bar.
A dot diagram can communicate the same twelve observations
A dot plot puts one dot at each numerical value for each observation. For Group A there is one dot at 6, one at 7, two at 8, one at 9, two at 10, and one each at 11, 12, 13, 14 and 24.
The large empty gap between 14 and 24 becomes visually obvious. If the axis is honestly scaled with equal numerical increments, readers can see why the range is so much larger than the central distribution.
Do not pack the axis between 14 and 24 into the same visual width as a one-minute interval without a clearly identified break or distortion. An attractive chart that hides real distances gives readers the wrong impression.
A simple histogram and a complete frequency check
Suppose the two fictional groups are displayed using equal-width five-minute intervals. For Group A, the interval 5 ≤ t < 10 contains five observations. The interval 10 ≤ t < 15 contains six. The interval 15 ≤ t < 20 contains zero, and 20 ≤ t < 25 contains one.
The frequencies 5 + 6 + 0 + 1 = 12 account for the complete sample. With equal-width intervals, the bar height can represent frequency directly; the bar boundaries should be consistent and the horizontal axis labelled in minutes.
Group B has four observations below ten minutes and eight from ten up to but not including fifteen. When comparing histograms, the same interval definitions and numerical scales prevent one group from looking artificially more variable.
What cumulative frequency adds to the interpretation
Cumulative frequency counts how many observations are below a specified upper interval boundary. For Group A, five observations take less than ten minutes, eleven take less than fifteen, and twelve take less than twenty-five.
For Group B, four observations are below ten minutes, while all twelve are below fifteen. This shows a difference that the equal median of ten minutes alone does not communicate.
A cumulative-frequency diagram accumulates counts as the boundary increases; its curve should not fall when based on cumulative counts. The axis labels and total number of observations remain essential to interpretation.
A small data-interpretation trap: percent of observations
How many members of Group A completed the task in less than ten minutes? Five out of twelve, giving 5/12 × 100% ≈ 41.7%. It is not 5% and not 50% simply because the graph uses a five-minute interval.
For Group B, four out of twelve completed in less than ten minutes, giving 4/12 = 1/3 ≈ 33.3%. These proportions are calculated from the same denominator because each group contains twelve fictional observations.
Do not translate these sample proportions into statements about all users of a real library. The two distributions were invented to teach mathematics, not gathered through representative sampling.
The effect of one large observation
Replace Group A’s 24-minute result with 14 minutes while keeping the other eleven observations unchanged. The total falls from 132 to 122 minutes, so the new mean is 122/12 ≈ 10.2 minutes to three significant figures.
The median remains 10 minutes. The range falls from 18 minutes to 14 − 6 = 8 minutes. Using the same median-of-halves convention, Q1 stays 8 and Q3 stays 12.5, so the IQR remains 4.5 minutes.
This comparison shows why multiple summaries are useful. The extreme value influences the mean and range substantially, while the median and IQR stay stable in this particular example.
When it may be reasonable to flag an unusual value
One widely used exploratory convention identifies observations beyond the boundaries Q1 − 1.5×IQR and Q3 + 1.5×IQR as potential outliers. This is a statistical diagnostic, not proof of bad data and not a formula to assume is required in every SEC question.
Using Group A’s Q1 = 8, Q3 = 12.5 and IQR = 4.5, the upper exploratory boundary is 12.5 + 1.5×4.5 = 19.25 minutes. The value 24 lies beyond it, so it would be flagged for closer investigation under this particular convention.
A flagged observation should not automatically be deleted. It could be genuine and informative. An honest analyst would examine how the value was measured and report any justified changes rather than remove a point merely because it makes a chart look untidy.
A box plot must disclose its whisker convention
A basic five-number-summary box plot can draw whiskers to the recorded minimum and maximum. Another plotting convention may stop the whisker at the most extreme non-flagged value and display unusual observations as individual points.
If a reader sees Group A’s 24-minute value displayed as a separate marker, they should look for the plotting convention before concluding the picture contradicts the written five-number summary.
Using one whisker convention for Group A and a different one for Group B without explanation would make the comparison unreliable. Consistent visual rules are part of fair mathematical communication.
An axis that starts above zero can exaggerate differences
Suppose a chart displays two group means of 10 and 11 minutes. An axis starting at zero will show a modest difference in column height. An axis starting at 9.8 can make the one-minute gap occupy almost the entire plotting area.
That does not make the numbers false, but it changes the visual impression. For bar charts, where lengths encode quantities, a truncated baseline can be especially misleading unless clearly justified and labelled.
Ask which information is most useful: the absolute mean difference of one minute, the full distributions, or how each group varies. A graphic can be technically legible while still encouraging an exaggerated conclusion.
The same box width can look different on unequal axes
If one box plot uses an axis from 0 to 30 minutes and another uses 6 to 15 minutes, the shapes cannot be compared by visual width alone. Group B could look visually wider even with the smaller IQR simply because its axis is more zoomed in.
To compare the fictional groups fairly, draw both box plots on the same numerical scale, or state the numerical summaries directly: both medians 10, Group A IQR 4.5 and Group B IQR 2.
Clear labels and transparent scales are not decorative details. They are mathematical choices that determine whether the visual accurately represents the data.
Why a sample size of twelve is a limitation
These invented groups contain only twelve observations each. A descriptive comparison of the given numbers is valid, but it cannot establish how future visitors to an actual library will behave or how an entire student population would perform.
Further observations might change a mean, a percentile or the apparent frequency of unusual values. A real study would need an explicit sampling method, comparable tasks and measurement rules.
Good mathematical writing uses phrases such as in this sample or under this model. It avoids unsupported claims about the people, school or place represented by an invented classroom exercise.
Correlation is not causation, even in a tidy chart
Suppose a fictional follow-up study found that readers who used a printed catalogue had shorter search times. The chart might show an association, but it would not alone establish that the catalogue caused the difference.
Readers could have been assigned different tasks, arrived with different prior knowledge or used the library under different conditions. Without further study design details, those possibilities remain unresolved.
A Mathematics student does not need to write a long research paper to respect this distinction. One sentence acknowledging that the data describe the observed pattern, not necessarily its cause, can make an interpretation more accurate.
The choice of chart should follow the reader’s question
Use a stem-and-leaf diagram or dot plot when exact individual whole-minute observations matter. Use a box plot when comparing medians and IQRs across groups. Use a histogram when the interest lies in frequencies across numerical intervals.
Use cumulative frequency when asking how many observations are below a threshold or when estimating a percentile from a drawn cumulative distribution. Use a table when readers need specific values without reconstructing them from a graph.
Each representation gives some information greater prominence and hides other detail. Before selecting the chart, ask: What specific question must the reader answer?
A tutor’s first-error diagnostic for a marked statistics paper
An incorrect median may result from unsorted data or choosing only one central value in an even-sized list. An incorrect IQR may come from mixing quartile conventions. An incorrect mean could be a frequency-count mistake rather than an addition mistake.
A misleading comparison may come from choosing the wrong spread measure or misreading two different axis scales. These errors have distinct remedies: reorder values, revisit quartile positions, recalculate frequencies or practise interpreting graphical conventions.
Instead of labelling the whole topic ‘statistics weak’, record the exact first unsound step and ask for a corrected explanation on a changed data set.
Independent practice with a new nine-score data set
Try the invented values 3, 4, 4, 6, 7, 8, 9, 9, 13. They sum to 63, so the mean is 7. With nine ordered observations, the fifth is the median, also 7. The range is 13 − 3 = 10.
Using the median-of-halves convention that excludes the central 7, the lower four values are 3, 4, 4, 6 and give Q1 = 4. The upper four are 8, 9, 9, 13 and give Q3 = 9; the IQR is 5.
Close the worked answers and solve a changed data set independently. A correct first attempt a few days later is more convincing than an immediate copy of the same calculation.
A seven-day revision plan for centre, spread and charts
Day 1: order observations and check the total frequency. Day 2: calculate means, medians, modes and ranges. Day 3: determine quartiles under a stated convention. Day 4: interpret box plots on a shared numerical scale.
Day 5: construct a stem-and-leaf diagram and frequency histogram. Day 6: compare cumulative-frequency thresholds and identify a potentially misleading display. Day 7: solve unseen data without a chapter heading naming the required statistic.
Keep a one-page error log with ordering, mean total, median position, quartile method, axis scale, IQR versus range, sampling limitation and unsupported causal claim as possible first-error categories.
Frequently asked questions about the invented library survey
Can the mean and median differ? Yes. Group A has mean 11 and median 10 minutes. Can two groups share the same median yet have different IQRs? Yes, Group A’s IQR is 4.5 and Group B’s is 2 under the stated convention.
Which group has the lower standard deviation? Group B, at roughly 1.08 minutes versus Group A’s 4.55. Does a flagged outlier prove an observation is incorrect? No; it indicates a value worth examining.
Should a histogram, box plot and dot plot all look the same? No. They emphasize different features of the same data. Does any result describe a real Bukit Timah library? No. Every observation is fictional educational material.
A practical three-pupil Mathematics tutorial in Bukit Timah
The original, unchanged eduKateSG Mathematics small-group reference describes premium tutorials of up to three pupils, with 1.5-hour weekly lessons, careful explanations, guided corrections and independent practice. Its stated venue is 8 Fourth Avenue, Singapore 268674, near Sixth Avenue MRT. This article does not advertise a newly opened branch or imply that a particular class has a vacancy.
A student who understands average but confuses quartiles needs a different correction from one who calculates accurately but makes unsupported claims about a misleading graph. A useful lesson begins with the learner’s own script, names the first wrong decision, reteaches that principle, and checks it on an altered data set after a gap.
Enquire about a parent–student Mathematics consultation with a recent independently completed statistics exercise and the learner’s exact examination year.
Use the right national Mathematics syllabus
The first Singapore–Cambridge SEC examinations take place in 2027, with G3 Mathematics code K310. Its official syllabus includes data presentation, mean and median, quartiles, percentiles, range, interquartile range and standard deviation. The SEAB subject list gives O-Level Mathematics 4052 as the earlier reference code. A learner sitting 2026 O-Level Mathematics should follow the applicable 2026 documents, not treat the 2027 SEC syllabus as identical.
Official sources and continuing reading
- SEAB: 2027 SEC G3 Mathematics K310 syllabus (PDF)
- SEAB: 2027 school-candidate G3 subject codes
- eduKateSG: original Mathematics small-group programme (unchanged)
- Statistics and Probability Clinic
- Box Plots, Median and IQR
- Histograms and Cumulative Frequency Diagrams
- School Canteen Cost, Revenue and Break-Even
- Cylinder Tank Flow and Time Graphs
- Mobile Plan Cost Graphs
All data and charts described here are original fictional teaching examples, not findings about real people or libraries. No grade or examination question is promised.