Small Group Tutorials

Here to help students catch up, keep up, and move ahead. Book a consultation here.

How to Perform in the new G3 SEC Examinations | Learner’s Guide Vol 1452 | Mathematics Library Return-Time Audit — Quartiles, Box Plots, Spread and Fair Decisions

Two groups test an imaginary school-library return track. One group reports that a moving block takes 0.74 seconds on average to travel down the track. Another group reports 0.92 seconds. It is tempting to stop there and declare the first surface the better choice. Mathematics asks us to look further. Are the trials comparable? How spread out are the times? What do the median, quartiles and standard deviation reveal? How would an unusual reading change the story? And what cannot be concluded even after every calculation is correct?

This learner’s guide takes those questions in order. It turns a simple Physics demonstration into a complete G3 Mathematics investigation involving ordered data, averages, ranges, quartiles, interquartile ranges, five-number summaries, box plots, frequency tables, cumulative frequency, standard deviation and constrained decisions. Every trial and device is fictional educational material. This is not a design or safety recommendation for an actual library return system.

Quick information: The 2027 G3 SEC Mathematics subject is K310, previously 4052. Both examination papers are 2 hours 15 minutes and 90 marks, weighted equally. Begin with a well-defined variable and unit; arrange raw observations in order; interpret centre and spread together; label quartile conventions; do not substitute an average travel time for actual processing capacity; state the sample limitation before making a general claim; finally, check that any proposed action follows from the evidence.

1. The SEC framework and why this is a Mathematics article

The Singapore-Cambridge Secondary Education Certificate begins in 2027, with G1, G2 and G3 subject levels under the new system. For standalone G3 Mathematics, the official 2027 K310 syllabus and SEAB subject list set the assessment framework. Paper 1 and Paper 2 each last 2 hours 15 minutes, each carry 90 marks, and each contribute 50% of the subject grade. Examination tasks include reasoning, applying techniques, interpreting information and mathematical problem solving, not only reproducing procedures.

The curriculum’s statistical tools allow candidates to interpret and compare distributions, measures of central tendency and dispersion, and graphical displays including box-and-whisker plots and cumulative-frequency representations. This guide is not an official sample paper. The library context is an invented way to rehearse these skills in meaningful order. To study the mechanics of the track itself, use Volume 1451: Physics Library Return-Chute Model. To understand why an integrated decision can be defeated by a faulty assumption, use Volume 1450: EMS Error-Detection Masterclass. This article has one canonical focus: what a data set permits a mathematically careful learner to conclude.

2. Set the question before collecting numbers

A data exercise begins with the measurement definition. Here the variable is the travel time, in seconds, for a teaching block to move from a specified release point to a fixed finish sensor. It is not the time a pupil spends returning a book, the wait between successive books, the time needed to clear the receiving tray, or the speed at the track exit. Those related quantities would require their own measurements and definitions.

Two deliberately controlled surface conditions, A and B, are compared using the same block, track geometry, release method and measurement positions. For this expanded statistics exercise, suppose we have eight new illustrative trial times for each condition. They are an extension of the earlier five-trial illustration, not a claim that the exact original readings were re-measured. The figures are chosen to permit transparent worked calculations and meaningful interpretation.

Surface A, seconds: 0.70, 0.72, 0.73, 0.74, 0.74, 0.75, 0.76, 0.78. Surface B, seconds: 0.88, 0.90, 0.91, 0.92, 0.92, 0.93, 0.94, 0.96. Both lists have eight measurements and are already in ascending order. The absence of extreme anomalies is a property of this teaching data, not a guarantee about actual trials.

3. Mean and median: the first useful comparison

For Surface A, the sum of the eight readings is 5.92 seconds, giving a mean of 5.92/8=0.74 seconds. For B, the sum is 7.36 seconds, so the mean is 7.36/8=0.92 seconds. The observed mean difference is 0.18 seconds. It is an average difference across these fictional trials, not an estimate of how much time a real school will save per returned book.

For eight ordered observations, the median is the average of the fourth and fifth values. Both central A readings are 0.74, so the A median is 0.74 seconds. Both central B readings are 0.92, so the B median is 0.92 seconds. In this example each median equals its respective mean. That coincidence is useful for teaching but is not generally guaranteed. It arises because of the constructed symmetry of these small distributions.

The mean uses the magnitude of every value. The median depends on order and the central position. If a reading is unusually large, the mean may move substantially while the median stays comparatively stable. Choosing which statistic to report is part of interpreting data; simply calculating every available average does not ensure that the conclusion is appropriate.

4. Range describes the full observed span

Range equals maximum minus minimum. For A it is 0.78−0.70=0.08 seconds; for B it is 0.96−0.88=0.08 seconds. The identical ranges do not say that the time distributions have identical locations. B’s times are larger throughout. A comparison should state a centre and a measure of spread, not select one statistic and pretend it tells the complete story.

The range is sensitive to extremes. Suppose a new A reading were 1.20 seconds because the apparatus was accidentally blocked. The new maximum would change the range dramatically. That observation should not be deleted merely because it is inconvenient, nor automatically assumed to reflect ordinary motion. First establish whether it was a valid reading, a changed condition, a measurement error or a genuine rare outcome. Statistical judgement begins with that investigation.

5. Quartiles: make the method explicit

Quartiles divide ordered data into portions. For this exercise we use the familiar median-of-halves convention for an even-sized list of eight values: the lower quartile Q1 is the median of the first four readings, and the upper quartile Q3 is the median of the last four. With A, Q1=(0.72+0.73)/2=0.725 seconds and Q3=(0.75+0.76)/2=0.755 seconds. With B, Q1=(0.90+0.91)/2=0.905 seconds and Q3=(0.93+0.94)/2=0.935 seconds.

The interquartile range is Q3−Q1. It is 0.030 seconds for both conditions. The IQR describes the spread of the central part of the ordered distribution, while the range describes the span from minimum to maximum. For some sample sizes, different accepted quartile calculation conventions can lead to slightly different results; follow the method required by the examination question and avoid silently switching definitions mid-solution.

Each group’s middle fifty percent is similarly concentrated in these teaching readings. The samples do not overlap in observed times: the maximum A reading, 0.78 seconds, is below the minimum B reading, 0.88 seconds. That is a strong descriptive fact about the supplied sixteen observations. It is still not proof that every possible future A trial will always be quicker than every future B trial.

6. The five-number summary

A box-and-whisker plot is built from a five-number summary: minimum, Q1, median, Q3 and maximum. For A the five values are 0.70, 0.725, 0.74, 0.755 and 0.78 seconds. For B the five values are 0.88, 0.905, 0.92, 0.935 and 0.96 seconds. A well-labelled diagram plots the two groups against the same time axis, making positions and spreads visually comparable.

The box runs from Q1 to Q3 with a line at the median. For the simple school box plots used in this practice, whiskers extend to the sample minimum and maximum. Some statistical software uses a different whisker convention based on outliers. That distinction matters if a student compares a software graph with a classroom five-number-summary graph. Define the convention rather than claiming every drawing with a box must use the same whisker endpoints.

Because every B observation is 0.18 seconds above the corresponding ordered A observation, B’s box plot is a translation of A’s to the right by 0.18 seconds. Its box length, whisker lengths, range and IQR stay the same. Translation changes the location of the distribution, not its spread. This is a powerful shortcut, but explain it from the data before invoking it.

7. Standard deviation measures another kind of spread

The G3 Mathematics study of data includes standard deviation as a measure of dispersion. For these eight observations, the mean of A is 0.74 seconds. The deviations are −0.04, −0.02, −0.01, 0, 0, +0.01, +0.02 and +0.04 seconds. Their squared values sum to 0.0042 square seconds. Under the population standard-deviation convention used for this full supplied data set, variance is 0.0042/8=0.000525 square seconds, and standard deviation is approximately 0.0229 seconds.

The B data are obtained by adding 0.18 seconds to every A value. Their differences from their new mean are therefore identical, and their standard deviation is also about 0.0229 seconds. Adding a constant to all measurements shifts the mean and median but does not change range, IQR or standard deviation. If every reading were instead multiplied by two, the standard deviation and other spread measures would scale by two. These are transformations of the data, not claims about physical causation.

Pay attention to whether a statistical formula divides by n or n−1. A sample standard deviation intended to estimate a population parameter may use n−1, while the descriptive population standard deviation of the complete supplied list uses n. The question, syllabus conventions and calculator mode determine the appropriate answer. A candidate can show excellent arithmetic but quote the wrong statistic by silently using the wrong calculator setting.

8. What makes a comparison fair?

Comparing A and B is sensible only because the basic conditions were defined consistently. If the test block’s mass changed for B, or the track length changed, an observed difference might not be attributed to surface condition alone. If the sensor triggered differently between groups, timing bias could shift an entire distribution. The mathematical calculations would still be valid descriptions of the recorded numbers, but the interpretation of the cause would be weaker.

A reliable comparison sentence joins what is shown with what is limited. For example: ‘In these eight controlled model trials per surface, A had a lower mean and median travel time than B, while both sets had the same range, IQR and population standard deviation. The readings support a difference under the tested conditions, not a universal claim about all possible equipment or operating situations.’ Such a sentence is more informative than ‘A is much better’.

9. From raw observations to a frequency table

Suppose we want to summarise a separate classroom batch of twenty travel-time readings, in seconds: 0.68, 0.70, 0.72, 0.73, 0.74, 0.74, 0.76, 0.77, 0.78, 0.79, 0.81, 0.82, 0.83, 0.84, 0.86, 0.88, 0.89, 0.91, 0.93 and 0.97. Here we use equal-width classes: 0.65 ≤ t < 0.75, 0.75 ≤ t < 0.85, 0.85 ≤ t < 0.95, and 0.95 ≤ t < 1.05. These twenty values form a separate fictional data set, not a combination of the A and B lists.

The class frequencies are 6, 8, 5 and 1 respectively, totalling 20. A useful self-check is to mark each reading into exactly one class, especially values at a boundary. A value of 0.75 belongs to the second class, not the first, because the first interval is closed on the left and open on the right. A frequency table is a rule for categorising observations, not a decorative summary.

With equal class widths of 0.10 seconds, the frequency histogram may show bar heights proportional to 6, 8, 5 and 1. With unequal widths, use frequency density to represent the data correctly. A tall bar based on a broad class can otherwise create a misleading impression. Read the axis label and class interval before inferring which range contains more observations.

10. Cumulative frequency: the running total

The cumulative frequencies at the upper class boundaries are 6, 14, 19 and 20. They answer how many recorded observations fall below 0.75, 0.85, 0.95 and 1.05 seconds respectively. The endpoints and inequality notation matter. The final cumulative total should match the sample size, 20. If it does not, inspect the frequency table for omission or duplication.

A cumulative-frequency curve can be used to estimate the median and quartiles, often by reading approximate positions at one quarter, one half and three quarters of the total count. These graphical estimates depend on class grouping and interpolation and need not match exact raw-data order statistics perfectly. Label them as estimates. Do not claim more digits of accuracy than the graph supports.

For the raw list of twenty readings, the exact median is the mean of the tenth and eleventh values, (0.79+0.81)/2=0.80 seconds. Under the median-of-halves convention, Q1 is the average of the fifth and sixth readings, both 0.74, hence Q1=0.74. Q3 is the average of the fifteenth and sixteenth readings, (0.86+0.88)/2=0.87. The exact IQR is 0.13 seconds. These raw-data values are useful checks on a graph-based estimate.

11. Percentiles and the meaning of ‘typical’

A percentile is a position within an ordered distribution; it is not automatically a percentage of time saved. Saying a timing lies at the 75th percentile means its position relative to the sample, under an explicit convention, is around the point below which about 75% of observations lie. Saying that it is ‘75% faster’ would be a wholly different statement and usually unsupported. Candidates should explain percentile questions in the variable’s units and the population being described.

A small sample may contain the observed median but still provide a poor estimate of a wider system’s long-term performance. Do not use the word ‘typical’ without asking what group, period and conditions are represented. The centre of a carefully selected demonstration sample is a fact about that demonstration, not a guarantee for thousands of uses.

12. An outlier is a question, not an excuse to erase data

Suppose Surface A produces an additional 1.20-second observation. It could have resulted from an obstruction, a changed release point, a timing fault or genuine variation. Recheck the record and apparatus conditions. An outlier may indicate an error, but it may also represent important behaviour. A responsible analysis may show results with and without a clearly identified invalid reading, giving reasons, instead of quietly removing a value to improve the mean.

The mean of the original eight A timings is 0.74 seconds. With 1.20 added as a ninth reading, the sum becomes 7.12 seconds and the mean is approximately 0.791 seconds. The median remains 0.74 seconds because it is the fifth ordered value. The disparity illustrates the mean’s sensitivity to an extreme observation. It does not by itself answer whether that ninth trial should be accepted.

13. Why average travel time is not service capacity

An important error in real-world interpretation is to divide one hour by average movement time and call the result a safe hourly handling rate. At 0.74 seconds per block, the arithmetic 3600/0.74 is roughly 4,865 journeys per hour, but that assumes a new item can start every 0.74 seconds with no handling, clearance, maintenance or interference. The data report travel times, not necessarily spacing between independent returns. That throughput conclusion is unsupported.

Suppose a separate hypothetical workflow requires 12 seconds of clearance per item. An idealised continuous one-hour limit would be 3600/12=300 items, if every interval can be used and no other restriction applies. If 10 minutes are unavailable, 50 minutes of operation would allow at most 50×60/12=250 items in the same simplified model. These are model capacities under stipulated conditions, not engineering approvals for actual physical machinery.

14. Decisions with multiple constraints

Imagine a committee wants a model that is quick, repeatable and suitable for further testing. Lower mean travel time would favour A in the recorded data. Equivalent spread measures would give neither condition a dispersion advantage for the eight readings. But equipment durability, safe handling, maintenance and operational compatibility have not been supplied. The correct mathematical decision might therefore be to prioritise A for a controlled follow-up trial while withholding a real operational recommendation.

The word ‘best’ depends on an objective. A faster result need not be safer, more reliable under prolonged use, cheaper or easier to repair. Before comparing options, ask what is being optimised and what constraints must also be satisfied. A strong problem-solving answer makes the objective function and domain explicit where the task supplies enough information.

15. Thirty-two graded practice decisions

Decision 01 — What variable was measured?

Worked answer: Travel time from fixed release to finish, in seconds. Reasoning: The unit and event definition prevent accidental substitution of waiting or handling time. Your next move: Explain how this differs from exit speed. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 02 — Which data set has the lower mean?

Worked answer: A, 0.74 s compared with B, 0.92 s. Reasoning: Use the sums of all eight readings divided by eight. Your next move: Show the correct mean expression for each. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 03 — What is the mean difference?

Worked answer: 0.18 s, found by subtracting 0.74 from 0.92. Reasoning: Difference is not automatically a percentage improvement. Your next move: State whether the reference value changes a relative comparison. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 04 — Find both medians.

Worked answer: A is 0.74 s; B is 0.92 s. Reasoning: Each has eight ordered observations, so average positions four and five. Your next move: Explain why an unordered list must first be sorted. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 05 — Why not infer identical variability from equal medians?

Worked answer: Equal medians concern centre, not spread; in this example the medians are not even equal. Reasoning: Every statistical comparison needs a correctly identified quantity. Your next move: Give an example of two lists with one median but different ranges. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 06 — What is A’s range?

Worked answer: 0.08 s, from 0.78−0.70. Reasoning: A range reports extreme observed span. Your next move: Identify the minimum and maximum first. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 07 — What is B’s range?

Worked answer: 0.08 s, from 0.96−0.88. Reasoning: Range does not report where the distribution is positioned. Your next move: Explain how both lists can have the same range but different medians. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 08 — Calculate A’s lower quartile.

Worked answer: Q1=0.725 s under the stated median-of-halves convention. Reasoning: It is the average of 0.72 and 0.73 in the lower four values. Your next move: State the convention when the method could otherwise be ambiguous. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 09 — Calculate A’s upper quartile.

Worked answer: Q3=0.755 s. Reasoning: Use 0.75 and 0.76 from the upper four observations. Your next move: Check that Q3 is no less than the median. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 10 — Calculate B’s quartiles.

Worked answer: Q1=0.905 s and Q3=0.935 s. Reasoning: Adding 0.18 to each A reading adds 0.18 to its quartiles. Your next move: Verify the shift using the raw values. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 11 — Compare interquartile ranges.

Worked answer: Both IQRs are 0.030 s. Reasoning: A shift changes location but not the spread of the central half. Your next move: Describe one situation where IQR is informative. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 12 — What does a five-number summary include?

Worked answer: Minimum, lower quartile, median, upper quartile and maximum. Reasoning: A box plot can be constructed from these five values. Your next move: Give A’s full five-number summary in ascending order. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 13 — Can the two box plots use different axes?

Worked answer: They should share a scale for fair visual comparison. Reasoning: Different scale choices can distort a viewer’s impression. Your next move: Describe how to label the horizontal seconds axis. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 14 — Why are B’s box and whiskers shifted?

Worked answer: Every ordered B reading is A plus 0.18 s. Reasoning: A constant translation shifts location while preserving differences. Your next move: Explain what happens to the median and IQR. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 15 — Does identical IQR mean identical data?

Worked answer: No. It indicates the same span between Q1 and Q3 under the chosen definition. Reasoning: Distributions can differ in position, shape and tails. Your next move: Invent two distributions with matching IQR but distinct medians. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 16 — What is A’s population standard deviation?

Worked answer: Approximately 0.0229 s for the eight supplied values. Reasoning: The sum of squared deviations is 0.0042 s²; divide by eight and take the square root. Your next move: Check the variance before rounding the standard deviation. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 17 — What is B’s standard deviation?

Worked answer: Approximately 0.0229 s under the same descriptive convention. Reasoning: Adding a constant preserves deviations from the shifted mean. Your next move: State why the units of SD remain seconds. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 18 — What would happen if every time doubled?

Worked answer: Mean, median, range, IQR and standard deviation would double. Reasoning: Scaling the variable by a positive factor scales location and spread accordingly. Your next move: Check one pair of values to verify. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 19 — Why does n versus n−1 matter?

Worked answer: They define different variance estimators or descriptive conventions. Reasoning: Calculator mode and the question’s required statistic determine the appropriate denominator. Your next move: Label the result you are quoting. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 20 — What does the frequency-table total equal?

Worked answer: 20 observations in the separate batch. Reasoning: Frequencies must count every reading exactly once. Your next move: Add 6, 8, 5 and 1. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 21 — Which frequency class contains 0.75?

Worked answer: The 0.75 to below-0.85 class. Reasoning: Intervals here are closed at the lower bound and open at the upper bound. Your next move: Explain which class contains 0.85. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 22 — Why do unequal histogram widths require care?

Worked answer: Frequency density allows bar area, not raw height alone, to represent frequency. Reasoning: Otherwise broad groups may look disproportionately important. Your next move: Describe how you would calculate frequency density. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 23 — What are the cumulative frequencies?

Worked answer: 6, 14, 19 and 20. Reasoning: Each entry is the running total through its boundary. Your next move: Label the upper boundary associated with each total. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 24 — What is the exact median of the twenty readings?

Worked answer: 0.80 s, the average of the tenth and eleventh values. Reasoning: An even sample has two central positions. Your next move: Identify their observed values. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 25 — What is that sample’s IQR?

Worked answer: 0.13 s under the specified median-of-halves rule. Reasoning: Q1=0.74 and Q3=0.87. Your next move: Explain why a graph read-off might be approximate. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 26 — Can we call the 75th percentile ‘75% faster’?

Worked answer: No. Percentile rank is a position within a distribution, not a percent change. Reasoning: The two phrases describe different mathematical objects. Your next move: Write a correct percentile explanation in seconds. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 27 — How does adding 1.20 s affect A’s mean?

Worked answer: It raises the mean from 0.74 to about 0.791 s for nine readings. Reasoning: An extreme observation can pull the arithmetic average. Your next move: Show the new total 7.12 divided by nine. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 28 — Does the added outlier change A’s median?

Worked answer: No; the ninth ordered observation set still has 0.74 as its fifth value. Reasoning: The median may be more resistant to an extreme value. Your next move: Explain why the outlier should not automatically be deleted. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 29 — Can 3600/0.74 prove hourly library capacity?

Worked answer: No. Travel time is not the same as clearance time or safe launch interval. Reasoning: A numerical quotient is not necessarily the requested real-world quantity. Your next move: Name a missing operational measurement. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 30 — At 12-second intervals, what is an ideal hourly limit?

Worked answer: 300 items per uninterrupted hour under that simple interval model. Reasoning: One hour contains 3600 seconds; divide by 12. Your next move: Give a practical reason the real count may be lower. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 31 — At 50 available minutes and 12-second intervals, what is the ideal maximum?

Worked answer: 250 items, assuming continuous use of all available time. Reasoning: Convert 50 minutes to 3000 seconds before division. Your next move: Check whether a start-up interval changes the interpretation. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

Decision 32 — How should the final recommendation read?

Worked answer: A is faster in these fictional controlled trials, with equal measured spread; further operational evidence is needed. Reasoning: A descriptive comparison is strong when it also states its limitations. Your next move: Write one factual conclusion and one justified next step. Before accepting the result, verify the unit, data set, statistic and question wording. If you give a numerical answer, finish with a brief interpretation that matches the scenario instead of leaving a number without meaning.

16. A complete worked response using the eight-trial data

Question: Compare the centre and spread of the travel-time readings for A and B. State one conclusion and one limitation. Worked solution: Mean and median are 0.74 s for A and 0.92 s for B. Both have range 0.08 s, interquartile range 0.030 s and population standard deviation about 0.0229 s. The lower centre of A indicates shorter recorded travel times under the defined conditions. The identical spread measures show equivalent dispersion in these examples. Since the figures come from only eight fictional controlled trials per group, the result cannot settle reliability or actual service performance in a school.

This worked answer names the group, centre, spread and limitation without claiming the faster group is more consistent. Many students lose a reasoning mark by writing ‘A has smaller standard deviation’ simply because it has smaller travel times. Absolute level and variability are different mathematical properties.

17. Diagnostic correction of three tempting mistakes

Mistake one: ‘Surface B is less consistent because its time is larger.’ Correction: greater mean or median does not imply larger spread. In the data, the two spreads are identical under the reported measures. Mistake two: ‘The middle half lasted 0.03 seconds.’ Correction: IQR measures the spread between Q1 and Q3; it does not mean the middle trials collectively lasted 0.03 seconds. Mistake three: ‘A can process 4,865 books each hour.’ Correction: movement time alone does not establish arrival spacing, queue dynamics or equipment clearance.

The repair technique is to identify the exact conversion that failed: centre to spread, statistical interval to duration, or observed transit time to operational capacity. A learner can know every formula but repeatedly make these category changes in word problems. Marking the first wrong interpretation is more helpful than simply supplying the correct final number.

18. A four-week statistical-reasoning plan

Week one — foundation: read tables, sort observations, calculate mean, median and range, and check whether the question asks for a time, count or speed. Use small lists and verify them by hand. Week two — distributions: calculate Q1, Q3 and IQR, produce five-number summaries, and draw parallel box-and-whisker plots on common axes. Explain every comparison in plain language.

Week three — grouped data: practise class boundaries, frequency density where class widths vary, cumulative-frequency curves, medians and quartile estimates. Mark whether an answer is exact or estimated. Week four — SEC transfer: complete mixed contexts under timed conditions, including outliers, standard deviation and capacity claims. Record whether lost marks arose from arithmetic, choosing a statistic, interpreting units or making an unsupported real-world inference.

Do not practise the same list of times until it feels familiar and mistake repetition for understanding. Replace the data with a shifted, scaled or unevenly distributed list. If the learner can explain which statistics change and why, the skill has begun to transfer. If they can calculate the mean but cannot say what the measurement represents, return to the first definition rather than adding harder formulas.

19. A family-friendly bridge from PSLE to secondary statistics

A Primary 6 pupil need not be expected to complete G3 SEC statistical questions immediately after PSLE. Posting Groups concern admission and entry arrangements, while G1, G2 and G3 describe subject-level study; progress depends on the individual learner’s readiness. The post-PSLE learning route gives a sensible foundation-first transition, and the Sengkang school-choice guide helps families understand their wider context.

If a student struggles with numerical comparisons, ask whether the problem is the four operations, fractions, interpretation of averages, reading axes or mathematical language. The Mathematics owner addresses that subject-specific foundation. The English owner is useful when a mathematically sound student misreads ‘at least’, ‘no more than’, ‘on average’ or ‘among respondents’. For the physical experiment, return to the Physics guide. The Science tuition owner concerns Primary 3–6 Science and should not be represented as a secondary Physics service.

20. Frequently asked questions

Why do a box plot when mean and median are known?

Mean and median describe centre, while a box plot shows the quartiles and observed extremes. It allows two distributions to be compared visually, especially where their centres look similar but their spread differs. It is not a substitute for recognising sample size or measurement conditions.

Is a smaller standard deviation always better?

Only if the aim is lower dispersion and the groups are meaningfully comparable. A low spread around a poor outcome may not be desirable. Clarify what variable is being measured and what decision the question asks the reader to make.

May I drop a reading that looks unusual?

Investigate it first. An identified measurement or recording error may warrant an explained correction. A genuinely extreme observation may be important evidence. Silent removal changes the data set and can distort the finding.

How precise should I make a graph-based quartile?

Give a result consistent with the scale and reading precision of the graph. A cumulative-frequency estimate may not match the exact quartile from ungrouped raw data. State or imply the appropriate degree of approximation rather than displaying unjustified digits.

Will the library scenario itself appear in the SEC paper?

No such prediction is made. The context is fictional practice for syllabus skills. Examination preparation is stronger when a learner can transfer the same statistical reasoning to unrelated tasks.

21. Official references and next reading

Use the 2027 SEAB K310 Mathematics syllabus for current examinable coverage and paper requirements, and the official G3 syllabus list to check codes and future updates. This volume teaches statistical comparisons and limits. For mechanics and experimental controls, see Volume 1451; for cross-subject error correction see Volume 1450. A useful final independent task is to make two new sets of eight times with the same mean but different IQRs, explain the comparison in two precise sentences, and identify what the data do not tell you about a real return system.