Singapore School Mathematics Operating Manual · Chapter 4
A calculator gives 14.4. A decision requires a value below 15. Is the condition definitely satisfied? Sometimes yes. Sometimes the answer depends on information that was lost when the inputs were rounded. The displayed result alone cannot tell us which situation we are in.
This chapter teaches how to turn rounded data into conclusions of the right strength. The goal is not another routine on decimal places. It is to distinguish guaranteed, possible and impossible outcomes when several exact values are compatible with the information supplied. A decision needs more than a plausible central estimate when the uncertainty could cross its boundary.
All dimensions, costs and journeys below are invented mathematical models. They are not specifications for construction, purchasing or safety decisions. Interval reasoning, dependence and some optimisation arguments are extensions beyond the requirements of particular school levels. Choose sections by prerequisite and consult the learner’s actual syllabus rather than treating every example as universal Primary, G1, G2 or G3 content.
Read a rounding interval · Test a threshold · Keep dependent quantities connected · Try decision problems · Read the worked answers
1. A rounded record represents more than one possible value
Suppose a positive length is reported as 8.6 cm correct to the nearest 0.1 cm. Using the positive-number round-half-up convention throughout this chapter, the true length L satisfies 8.55 ≤ L < 8.65. The lower midpoint rounds up to 8.6; the upper midpoint rounds up to 8.7 and is excluded.
The recorded value 8.6 is one value inside that interval, not a promise that the true measurement is exactly 8.600000 cm. A length of 8.56 cm and a length of 8.64 cm would both produce the same report.
Other explicitly specified rounding conventions can treat midpoint cases differently. Read the convention when it matters. This chapter uses positive half-up rounding so that each example has a clear and consistent endpoint rule.
The maximum absolute rounding error under this convention can equal half the rounding unit at an included midpoint. It is therefore misleading to write that the error is always strictly less than half a unit. For the report above, the error can be 0.05 cm when L = 8.55 cm.
For the foundational conversion and accuracy skills, use BTT’s Units, Scale and Measurement guide. Here we take the interval as the starting information and ask what a decision can legitimately conclude from it.
2. Three verdicts for a numerical condition
Keep 8.55 ≤ L < 8.65. Is L less than 8.7 cm? Yes, for every allowed value. The condition is guaranteed. No additional decimal place is needed to establish that particular conclusion.
Is L at least 8.6 cm? That is possible but not guaranteed. The value 8.64 satisfies it, while 8.56 does not. Both values are compatible with the same rounded report. A single yes-or-no answer would claim more than the data establish.
Is L at least 8.65 cm? That is impossible under the stated rounding model. The interval excludes 8.65 and everything above it. The qualifier matters: we are reasoning from the stated report and convention, not guaranteeing that a physical instrument or transcription was error-free.
“Possible” includes at least one allowed case. “Guaranteed” means every allowed case. “Impossible” means no allowed case. When a statement is possible but not guaranteed, its opposite is also possible. Two compatible examples on opposite sides of the threshold make that uncertainty visible.
These words describe the strength of the conclusion, not the student’s confidence or the calculator’s precision. A carefully justified “not determined” can be a stronger mathematical answer than an unjustified definite claim.
3. Strict and inclusive boundaries change the conclusion
For the same interval, L < 8.65 is guaranteed, but L < 8.64 is not. The value 8.645 is still allowed and exceeds 8.64. Rounding the upper boundary down to two decimal places before testing the condition would corrupt the reasoning.
Likewise, L ≥ 8.55 is guaranteed, but L > 8.55 is not, because the lower endpoint is included. The difference between “greater than” and “at least” is not merely a notation preference at a boundary.
An upper bound need not be a greatest possible value. The interval L < 8.65 has no largest allowed real value: any candidate below 8.65 can be replaced by a slightly larger allowed value. The number 8.65 is its least upper bound, not an attained maximum.
In school working, people sometimes use “maximum” informally for an upper bound. When the task concerns a guarantee exactly at the boundary, retain the distinction. State the interval and whether equality is permitted rather than relying on an ambiguous word.
For a younger learner, the same idea can be explored with whole numbers. A count less than 10 has greatest integer value 9. A real measurement less than 10 has no greatest allowed value. The type of quantity determines whether a boundary is attained.
4. Comparing two rounded measurements
An invented panel width P is reported as 120.0 cm to the nearest 0.1 cm. An opening width O is reported as 120.1 cm to the same precision. Their intervals are 119.95 ≤ P < 120.05 and 120.05 ≤ O < 120.15.
Under this simplified one-dimensional model, P is always less than O. The largest boundary for P is excluded, while the smallest allowed O is 120.05. Thus the rounded reports establish a strict ordering even though neither exact width is known.
They do not establish a clearance of at least 0.1 cm. For example, P = 120.049 and O = 120.05 give a gap of only 0.001 cm. The nominal difference 120.1 − 120.0 = 0.1 is not a guaranteed minimum gap.
If both widths were reported as 120.0 cm, the ordering would be uncertain. P = 119.96 and O = 120.04 make the opening wider. Reversing those values makes the panel wider. Equal values are also possible. Equal rounded reports do not prove equal exact measurements.
This is only a mathematical interval comparison. Actual fitting can depend on shape, alignment, tolerances, measurement error and required clearance. A classroom conclusion about one-dimensional widths should not be presented as an instruction to manufacture or install an object.
5. Addition and subtraction: choose the extremes for the target
Two positive lengths are independently allowed by the reports A = 12.4 cm and B = 7.2 cm, each to the nearest 0.1 cm. Thus 12.35 ≤ A < 12.45 and 7.15 ≤ B < 7.25. Here “independently allowed” means that no additional relationship between the errors has been supplied; it is not a probability assumption.
For the sum, the smallest possible value uses both lower endpoints: A + B ≥ 19.50. The upper boundary is 12.45 + 7.25 = 19.70, excluded. Hence 19.50 ≤ A + B < 19.70.
For the difference A − B, making A small and B large makes the result small. The lower boundary is 12.35 − 7.25 = 5.10, but it is not attained because B cannot equal 7.25. The upper boundary is 12.45 − 7.15 = 5.30, also not attained. Therefore 5.10 < A − B < 5.30.
The nominal difference is 5.2 cm. A requirement that the difference exceed 5.0 cm is guaranteed. A requirement that it exceed 5.25 cm is possible but not guaranteed. A requirement that it reach 5.30 cm is impossible within these intervals.
Do not apply a single rule such as “use all lower bounds for the lower answer”. Subtraction reverses the role of the subtracted quantity. The target expression determines which inputs make it increase or decrease.
6. Products: an area decision from rounded sides
A rectangular sheet has reported dimensions 1.20 m and 0.80 m, each correct to the nearest 0.01 m. The length l satisfies 1.195 ≤ l < 1.205, and the width w satisfies 0.795 ≤ w < 0.805. Assume no further relationship between them.
Both dimensions are positive, so their product increases when either dimension increases. The smallest area is 1.195 × 0.795 = 0.950025 m². The upper boundary is 1.205 × 0.805 = 0.970025 m², excluded. Thus 0.950025 ≤ A < 0.970025 m².
Is the area at least 0.95 m²? Yes. The smallest allowed area already exceeds 0.95. Is it at least 0.96 m²? Not guaranteed. The nominal product is exactly 0.96, but allowed lower dimensions produce less than that, while allowed higher dimensions produce more.
For explicit counterexamples, l = 1.196 and w = 0.796 give 0.952016 m², below 0.96. The pair l = 1.204 and w = 0.804 gives 0.968016 m², above 0.96. Both pairs round to the reported dimensions.
The extra decimal digits in the products are not a claim that the physical sheet was measured to six decimal places. They describe exact boundaries of the mathematical interval model. Interpretation of input precision and exact calculation with the interval endpoints are different jobs.
7. Division: a speed bound uses opposite input extremes
An invented journey distance is reported as 4.8 km to the nearest 0.1 km, and elapsed time as 20 minutes to the nearest minute. Let the exact values be D and T. Then 4.75 ≤ D < 4.85 and 19.5 ≤ T < 20.5.
The average speed in kilometres per hour is v = 60D/T. Because the quantities are positive, speed increases with distance and decreases with elapsed time. The lower boundary therefore uses the smallest distance and the largest time boundary. The upper boundary uses the largest distance boundary and the smallest time.
This gives 60(4.75)/20.5 < v < 60(4.85)/19.5. In exact fractional form, 570/41 < v < 194/13 km/h. Numerically, the boundaries are approximately 13.9024 and 14.9231 km/h. Neither is attained under the given endpoint rules.
Therefore v < 15 km/h is guaranteed in this model. The nominal calculation 60(4.8)/20 = 14.4 is not itself the proof; the entire allowed interval lies below 15. A claim that v < 14.4 is not guaranteed because values on both sides of 14.4 are possible.
The calculation concerns average speed over the whole journey. It does not bound the speed at every instant. A journey can contain faster and slower sections while retaining the same total distance and time. A correct interval must still be attached to the correct mathematical quantity.
8. Whole-number decisions can jump when a threshold is crossed
A modelled liquid quantity is reported as 30 litres to the nearest litre, so 29.5 ≤ V < 30.5 litres. Each idealised container holds at most exactly 10 litres, and every part of the liquid must be stored. Partly filled containers are allowed.
The nominal calculation suggests three containers. But an allowed quantity such as 30.4 litres needs four, while 29.6 litres needs three. The minimum number required by the unknown actual quantity is therefore not uniquely determined.
Four containers are sufficient for every allowed quantity because their total capacity is 40 litres. Three are sufficient for some allowed quantities but not all. “Minimum needed for the actual quantity” and “minimum that guarantees enough capacity despite the uncertainty” are different questions.
Now suppose the quantity is 29 litres to the nearest litre. Then 28.5 ≤ V < 29.5, entirely between 20 and 30. Exactly three containers are required for every allowed quantity. The uncertainty remains, but it no longer crosses a capacity threshold.
Discrete decisions are sensitive to boundaries because their answers change in jumps. A small change in a volume or attendance estimate may leave the required count unchanged for a while, then increase it by one when a boundary is crossed.
9. Correct the record before calculating its interval
Consider a different record written as “29.8 litres to the nearest whole litre”. Under ordinary whole-litre rounding, that is inconsistent: the rounded report should be a whole number. A careful reader should notice the mismatch before applying an interval formula.
The writer may have intended the nearest 0.1 litre, or may have copied the recorded number incorrectly. Those are different corrections. Without knowing which was intended, a reader should not silently choose a convenient interpretation and call its interval established information.
If the intended instruction were confirmed as the nearest 0.1 litre, the interval for the 29.8 report would be 29.75 ≤ V < 29.85. Three ten-litre containers would then suffice throughout the interval. But that conclusion depends on the corrected instruction, not on the inconsistent original sentence.
Interval reasoning cannot rescue a badly specified input. First determine whether the record, unit and rounding instruction are compatible. Only then derive the allowed values. A formula applied to an incoherent description can produce neat arithmetic without a justified model.
The general lesson is useful beyond rounding: validate the representation before using the result. A contradictory input should be corrected or clarified, not quietly transformed into a more convenient problem.
10. A budget can have a guaranteed count and a larger possible count
In an invented purchasing exercise, a budget is exactly $100, and the only uncertain quantity is the price of one identical item. The reported unit price is $7.70 to the nearest $0.10, so the price p lies in 7.65 ≤ p < 7.75 dollars. Any actual prices used as examples below are in whole cents.
Twelve items always fit the budget because 12p < 12(7.75) = $93. Thirteen items may fit: at p = $7.68, the total is $99.84. But at p = $7.72, the total is $100.36, above the budget. Both prices have the same stated rounded report.
Fourteen items cannot fit, since even the lowest allowed price gives 14(7.65) = $107.10. The actual maximum affordable count is therefore either 12 or 13. Twelve is the largest count guaranteed affordable; thirteen is the largest count that is possibly affordable.
The critical price for thirteen items is 100/13 dollars, approximately $7.6923. Prices at or below that threshold satisfy the budget; higher prices do not. A more precise actual price would settle which count is available.
This is a hypothetical arithmetic model, not advice about real prices or payment rounding. Real invoices may include conditions that the model intentionally omits. Its teaching purpose is to distinguish the count supported by every allowed price from the count supported by only some.
11. Do not vary the same unknown independently twice
Suppose 9.5 ≤ x < 10.5. What are the possible values of x − x? The answer is exactly zero. Treating the first x as 9.5 and the second as 10.49 would use two different values for the same unknown in the same expression.
A mechanical interval subtraction would give a broad enclosure between −1 and 1 if it forgot that the two occurrences are identical. That enclosure contains the true answer, but it is not the sharp set of possible values. The dependence makes the expression much more constrained.
The expression 2x − x simplifies to x, so its range is the original interval. Again, simplifying before treating repeated occurrences as separate inputs preserves information that a naive endpoint calculation would discard.
The same issue appears in geometry. If a rectangle has fixed perimeter 40 cm, its side lengths x and 20 − x are connected. One side cannot independently take its largest value while the other also takes its largest value. Their sum must remain 20.
Before combining intervals, ask whether the quantities are separate unknowns, repeated versions of the same unknown, or related through a constraint. The correct uncertainty calculation must preserve those relationships.
12. A fixed perimeter changes the area bounds
Let a rectangle have exact perimeter 40 cm and one side x reported as 10 cm to the nearest centimetre. Then 9.5 ≤ x < 10.5, and the other side is 20 − x. The area is A = x(20 − x).
Complete the square: A = 100 − (x − 10)². Over the allowed interval, the squared term lies from 0 to 0.25 inclusive. Zero occurs at x = 10. The value 0.25 occurs at the included endpoint x = 9.5, even though the other endpoint x = 10.5 is excluded.
Therefore 99.75 ≤ A ≤ 100 cm². Both bounds are attained: a 9.5 cm by 10.5 cm rectangle has area 99.75, and a 10 cm by 10 cm square has area 100.
It would be wrong to multiply independent side intervals and conclude that area could be as low as 9.5 × 9.5. Those two sides would sum to 19, violating the exact perimeter. The constraint excludes the supposed worst-case combination.
This example requires algebraic completion of the square and is an extension for learners with that prerequisite. Its central lesson is broader: a conservative calculation may be needlessly loose if it forgets a relationship, and a falsely precise calculation may be invalid if it invents one.
13. Rounded counts still have to be integers
An attendance count is reported as 50 to the nearest 10 people. Under the stated convention, its underlying value N satisfies 45 ≤ N < 55. Because N counts people, the allowed values are the integers 45 through 54.
There are not infinitely many admissible attendance values. A real-number interval is useful working, but the final domain restricts it. A value such as 50.4 people is not an alternative count.
Would 50 seats be enough? That is possible but not guaranteed: attendance could be 49 or 54. Would 54 seats be enough? Yes, because 54 is the largest admissible integer count. The real upper boundary 55 is excluded.
Now suppose the mean size of four classes is reported as 21 pupils to the nearest whole pupil. The exact mean lies in 20.5 ≤ m < 21.5. The total N = 4m therefore satisfies 82 ≤ N < 86, so possible integer totals are 82, 83, 84 and 85.
This does not mean each class has between 20.5 and 21.5 pupils. The report concerns a mean. Domain and quantity identity must both survive the interval transformation.
14. Percentage change depends on both the old and new values
An old measurement is reported as 60 units and a new measurement as 66 units, both to the nearest whole unit. Let their exact values be A and B, with 59.5 ≤ A < 60.5 and 65.5 ≤ B < 66.5. The percentage increase is 100(B/A − 1).
The nominal values give 10%. But the actual percentage increase can be lower or higher. A = 60.4 and B = 65.6 give approximately 8.61%. A = 59.6 and B = 66.4 give approximately 11.4%. Each pair produces the same rounded reports.
Since the new value is always above the old one, an increase is guaranteed. An increase of at least 10% is not. The claim about direction is stronger than uncertainty about the exact percentage might initially suggest.
For positive A and B, the lower percentage boundary uses the smallest B and largest A boundary: 100(65.5/60.5 − 1) = 1000/121%, approximately 8.26446%. The upper boundary is 100(66.5/59.5 − 1) = 1400/119%, approximately 11.7647%. Neither boundary is attained with these half-open intervals.
The denominator must remain the original value A. Dividing the change by B would calculate a different relative comparison. Uncertainty handling does not replace the need to choose the correct percentage base.
15. Numerical rounding is not the same as every other uncertainty
A rounding interval tells us which exact numerical values would produce a recorded value under a stated rounding rule. It does not automatically include instrument bias, transcription mistakes, changing conditions or an invalid model.
Similarly, a sample proportion rounded to 0.60 does not by itself prove that the population probability lies between 0.595 and 0.605. That interval concerns the unrounded sample proportion if it was rounded to the nearest 0.01. Sampling uncertainty about the population is a different question.
A calculator displaying many digits cannot remove uncertainty that entered through the measurement or sampling process. Nor does replacing a decimal with a symbolic expression create missing evidence. Exact arithmetic can faithfully transform uncertain inputs without making them certain.
Keep the layers separate: what was measured or counted, how it was recorded, what model connects it to the target, and what conclusion the result is being used to support. Each layer can introduce a different limitation.
The existing exact-answer and decimal guide owns the general representation choice. This chapter’s decision framework begins when the allowed input values, and the assumptions defining them, are clear.
16. Independent practice: what is actually guaranteed?
Use positive round-half-up rounding. Unless stated otherwise, quantities have no additional relationship beyond the supplied conditions. Give an interval or compatible examples to justify each conclusion.
1. A length is 6.2 cm to the nearest 0.1 cm. State its interval. Is it necessarily less than 6.25 cm? Is it necessarily greater than 6.15 cm?
2. Two lengths are reported as 4.3 cm and 2.1 cm, each to the nearest 0.1 cm. Bound their sum.
3. With the same data, bound the first length minus the second. Is the difference necessarily greater than 2.0 cm?
4. A rectangle has dimensions 5 cm and 3 cm, each to the nearest centimetre. Bound its area. Is its area necessarily at least 15 cm²?
5. A count is reported as 80 to the nearest 10. List the possible counts. What is the smallest seat capacity guaranteed sufficient?
6. An exact budget is $50. An item’s exact price is unknown but satisfies $4.90 ≤ p < $5.10. Are ten items guaranteed affordable? What count is guaranteed affordable?
7. A positive quantity x lies in 2.5 ≤ x < 3.5. Find the possible values of x − x and of 3x − 2x.
8. A positive radius is established exactly as √50 cm. A student reports it as 7.07 cm. What exact circle area should be used in a later part?
9. A volume is 40 litres to the nearest litre. Containers hold exactly 10 litres each. Is the actual minimum container count determined? What count guarantees enough capacity?
10. A rectangle has exact perimeter 24 cm and one side x satisfying 5.5 ≤ x < 6.5. Find sharp area bounds.
11. The mean of three integer counts is reported as 10 to the nearest whole number. Find the possible integer totals.
12. A recorded length is written as “7.4 cm to the nearest whole centimetre”. What should be checked before using an interval?
17. Worked answers and decision logic
1. The interval is 6.15 ≤ L < 6.25 cm. Therefore L < 6.25 is guaranteed. The claim L > 6.15 is not guaranteed because L = 6.15 is an allowed midpoint that rounds to 6.2. The inclusive version L ≥ 6.15 is guaranteed.
2. The inputs satisfy 4.25 ≤ A < 4.35 and 2.05 ≤ B < 2.15. Addition gives 6.30 ≤ A + B < 6.50 cm. The lower value is attained when both inputs are at their included lower endpoints. The upper boundary is excluded.
3. Subtract the largest second boundary from the smallest first value, and conversely: 4.25 − 2.15 < A − B < 4.35 − 2.05. Hence 2.10 < A − B < 2.30 cm. Both boundaries are excluded. The difference is necessarily greater than 2.0 cm because every allowed difference is already greater than 2.10.
4. The side intervals are 4.5 ≤ l < 5.5 and 2.5 ≤ w < 3.5. Both are positive, so 11.25 ≤ lw < 19.25 cm². An area of at least 15 is not guaranteed. For example, the allowed dimensions 4.6 and 2.6 give 11.96 cm², while 5.4 and 3.4 give 18.36 cm². The same reports support both sides of the threshold.
5. The real interval is 75 ≤ N < 85. The possible integer counts are 75, 76, 77, 78, 79, 80, 81, 82, 83 and 84. A capacity of 84 seats is the smallest guaranteed sufficient. A capacity of 83 would fail for the allowed count 84.
6. Ten items are not guaranteed affordable: p = $5.05 gives a total of $50.50. They are possible at p = $4.95, giving $49.50. Nine are guaranteed affordable because 9p < $45.90, below $50. Eleven are impossible within the budget because 11p ≥ $53.90. Thus nine is the largest guaranteed count.
7. The expression x − x is exactly zero. The expression 3x − 2x equals x and therefore lies in 2.5 ≤ 3x − 2x < 3.5. Do not assign different values to repeated occurrences of the same x to manufacture wider extremes.
8. The exact area is π(√50)² = 50π cm². The rounded report 7.07 is not the best available input for the later exact calculation. This example concerns preserving an established exact result, unlike a radius that was originally measured only to the nearest 0.01 cm.
9. The volume lies in 39.5 ≤ V < 40.5 litres. Four containers suffice for V ≤ 40, but five are needed for V > 40. Both cases are allowed, so the actual minimum is not determined. Five containers are the minimum count that guarantees adequate capacity across all allowed volumes.
10. The other side is 12 − x, so A = x(12 − x) = 36 − (x − 6)². The squared term ranges from 0 to 0.25 inclusive. Therefore 35.75 ≤ A ≤ 36 cm². The maximum is attained at x = 6 and the minimum at x = 5.5. Treating the two sides independently would ignore the exact perimeter.
11. The exact mean m satisfies 9.5 ≤ m < 10.5. Multiplying by three gives 28.5 ≤ N < 31.5. The possible integer totals are 29, 30 and 31. The interval is for the total after multiplication, not a statement that each separate count has to be near 10.
12. The report and rounding instruction are inconsistent under ordinary whole-centimetre rounding, which should produce a whole-number record. Check whether the intended precision was the nearest 0.1 cm or whether the recorded value is wrong. Do not silently choose one correction and present its interval as established information.
18. A useful decision record
For a difficult case, write four short lines: recorded data; allowed exact values; target relationship; conclusion strength. For example: “distance 4.8 km and time 20 minutes; intervals as stated; average speed 60D/T; definitely below 15 km/h within this model.”
If the threshold lies inside the allowed output range, supply one compatible case on each side. This both justifies uncertainty and reveals what extra measurement precision would settle the decision. Sometimes only one input needs better precision because the other contributes little to the disputed boundary.
Do not respond to every uncertain input by demanding more digits. A guarantee may already be available when the whole interval lies safely on one side. Conversely, a central value extremely close to a threshold may remain inconclusive even after a modest increase in precision.
The right next measurement is the one that can change the decision, not automatically the one with the longest decimal display. That is a mathematical information question before it becomes a calculation question.
When would another decimal place actually help?
Suppose a positive length is reported as 10.0 cm to the nearest 0.1 cm, and a comparison requires L < 10.02 cm. The interval 9.95 ≤ L < 10.05 crosses the threshold, so the report does not settle the comparison. More precise information might help, but its usefulness depends on the new report.
If a reliable second reading establishes 10.00 cm to the nearest 0.01 cm under the same idealised rounding model, then 9.995 ≤ L < 10.005. Every allowed value is below 10.02, so the condition becomes guaranteed. The new precision matters because it moves the entire allowed interval to one side of the decision boundary.
If the second report is instead 10.02 cm to the nearest 0.01 cm, the interval is 10.015 ≤ L < 10.025. It still crosses 10.02. The extra digit has narrowed the uncertainty without settling the strict comparison. A more detailed record is not automatically a decisive record.
There is also no need for another reading when the original interval already answers the actual question. The report 10.0 cm to the nearest 0.1 cm is enough to establish L < 10.1. Demanding more precision for that particular threshold adds work without changing the conclusion.
This reasoning is about an idealised numerical report, not a promise that repeated physical readings share the same accuracy or lack bias. Within the stated model, the decision test remains simple: identify the allowed set, locate the threshold, and ask whether any admissible values remain on both sides.
A useful written conclusion names both its strength and its assumptions: “Under the stated rounding rule, every compatible value is below the threshold.” When both outcomes remain possible, supply examples and say which missing information would distinguish them. This keeps uncertainty precise rather than vague.
19. Continue through the School Mathematics manual
Return to the BTT School Mathematics Operating Manual. Use Command Words, Answer Forms and Task Contracts to identify the required conclusion, Data Sufficiency, Unique Answers and Counterexamples to test whether the information settles it, and Linked Question Parts, Hence and Result Handoffs to preserve trustworthy values between parts.
For earlier prerequisites, use Units, Scale and Measurement and Inequalities, Intervals and Regions. The main Mathematics Hub retains the wider curriculum, tuition, diagnosis and examination routes.
20. Sources and boundaries
The interval constructions and decision problems are original. They deliberately distinguish basic rounding from further reasoning about thresholds, dependence and optimisation. Inclusion in this chapter is not a claim that a particular extension is prescribed for every school or subject level.
For formal coverage and examination instructions, consult MOE’s secondary curriculum information and the applicable SEAB document. The 2026 PSLE formats, 2027 SEC G3 syllabuses, and 2026 A-Level syllabuses were checked as distinct cohort doorways on 6 September 2026.
Precision is not the number of digits a student can produce. It is knowing which conclusions those digits, their inputs and their assumptions can support.