The immediate answer: identify the whole before you calculate
If a question says that the proportion of organisms with a characteristic rose, first find the denominator used at each time point. The proportion is the number with the characteristic divided by the whole group at that point. Two percentages cannot be converted into a comparison of absolute counts without the group totals. The most useful examination habit is to write the three quantities together: part, whole and part divided by whole.
This guide is about the Mathematics of changing population records, not about diagnosing a disease or selecting medicine. Its fictional numbers are designed to expose denominator errors, weighted-average mistakes, impossible integer answers and misleading percentage claims. Volume 0886, Biology, owns the explanation of inherited variation and selection. Here, the question is different: what do the counts permit us to calculate, and how do we know whether a numerical claim is valid?
An efficient G3 response defines the part and the whole, sets out the calculation, writes a unit or percentage label and checks whether the answer is possible. The final sentence should not confuse a difference of percentage points with a percentage increase. When the data are rounded, give an honest interval rather than inventing certainty. When the unknown is a count, remember that it must ordinarily be a whole number.
The 2027 examination context
From 2027 the Singapore–Cambridge Secondary Education Certificate replaces the previous GCE N- and O-Level examination names, with subjects examined at G1, G2 or G3 levels under Full Subject-Based Banding. The official SEAB 2027 school-candidate G3 listing, updated 26 October 2026, gives Mathematics code K310, English Language K300 and Biology K325; Biology combinations in Combined Science use the appropriate K327 or K328 designation. Always verify your own registered subject and examination year.
The 2027 G3 Mathematics K310 syllabus consists of two 90-mark papers, each worth half the assessment and lasting two hours fifteen minutes. Paper 2 includes extended problem-solving in a real-world context. This population-data clinic is an original worked application of percentages, ratios, rates, equations, interpretation and justified decisions, not an official SEAB paper. Consult the SEAB 2027 G3 index and SEC information page for official particulars.
Mathematics marks are not awarded for scientific-sounding commentary. A Mathematics answer should show the relevant variables, working, numerical result and interpretation. The Biology owner explains why a population may change; Mathematics tells us whether the measured change has been calculated and described correctly. The disciplines support each other while retaining separate assessment standards.
Use one stable notation
Write R for the recorded number with the specified characteristic, N for the number without it and T for the total. Then T = R + N and the recorded fraction is R/T. Multiply this fraction by 100% for the percentage. Where the question supplies rates per hundred examined, the same fraction is being scaled: 100R/T. You do not need advanced algebra to understand it, but clear notation becomes valuable when a problem gives incomplete totals.
Do not use the same letter for two different time points. A simple table with columns ‘earlier’ and ‘later’ and rows R, N, T and R/T keeps the problem under control. If the fraction changes, your table will show whether R increased, N decreased or both. When answering a longer question, leave the table visible rather than erasing it after one calculation. It becomes an audit trail for the final decision.
Students arriving from the PSLE Foundation Mathematics understanding route can make the first transfer by using part-whole bars. Put R and N inside a bar representing T. The symbols then express what the diagram already means. You can move from a bar model to algebra without changing the underlying relationship.
Worked example: a fraction rises while the total shrinks
At Time A, a fictional record contains R = 100 and N = 900. T is 1,000 and the proportion is 100/1,000 = 10%. At Time B, R = 120 and N = 180. T is 300, and the proportion is 120/300 = 40%. The proportion rises from 10% to 40%, which is an increase of 30 percentage points. The resistant count rises from 100 to 120, which is an increase of 20 people or organisms in the model, and a 20% relative increase in the count.
The total population, however, changes from 1,000 to 300: a fall of 700, or 70%. All those results are true at once. It would be false to say the number with the trait quadrupled just because the percentage increased fourfold. The fraction 40% is four times 10%, but R increased only from 100 to 120. The denominator explains the difference.
One clean answer sentence is: ‘The recorded fraction with the trait rose by 30 percentage points, while the absolute number rose by 20% and the overall sample total fell by 70%.’ This states three different measures without treating them as synonyms. Never write ‘increased by 30%’ when you really mean 30 percentage points.
Percentage points are not percentages
A percentage-point change is the subtraction of percentages. A relative percentage change is the difference divided by the starting value, then multiplied by 100%. If a fraction changes from 10% to 40%, the percentage-point change is 40 − 10 = 30 points. The relative percentage increase in the fraction is (40 − 10)/10 × 100% = 300%. That is correct mathematically, though often less useful than clearly stating the before-and-after fractions.
Consider the phrase ‘increased by 25%’. If a fraction begins at 20%, a 25% relative increase gives 20% × 1.25 = 25%. An increase of 25 percentage points would instead give 45%. The difference is enormous. A student who substitutes one interpretation for the other may carry a wrong total through the rest of the question.
When an exam question gives words without an explicit symbol, use the ordinary meaning of the phrase in context and show the calculation. If the wording specifically says ‘percentage-point increase’, subtract the two percentages. If it says ‘percentage increase in the number’, use the count as the base. Write the base clearly; this can prevent several follow-on mistakes.
Reversing a percentage to recover a count
Suppose a record states that 18% of 250 fictional entries have the trait. R = 0.18 × 250 = 45, and N = 250 − 45 = 205. The answer is a whole count, consistent with the statement. If 18% is an exact fraction of 250, 45 is the exact count. A student may mistakenly divide 250 by 0.18, producing a number far larger than the total, because they have confused which quantity is known.
Now reverse a different question: 45 entries are 18% of the total. Then 45/T = 0.18, so T = 45/0.18 = 250. Both tasks use the same equation; they differ in the unknown. Before calculating, write the relationship R = pT and mark which one of R, p and T is missing.
Check by substitution. If you claim T = 250, compute 45/250 = 0.18. An answer that fails its own original condition should not be submitted. That reverse check is particularly useful in a multi-part Paper 2 question where an early wrong value can contaminate later decisions.
Whole-number constraints are part of the problem
An exact statement that 17% of 230 entries carry a trait gives 0.17 × 230 = 39.1. But a discrete count cannot be 39.1 entries. This signals that at least one assumption is wrong: perhaps the percentage was rounded, perhaps the denominator refers to a different group, or perhaps the numbers are intentionally inconsistent and the examiner wants you to recognise it.
Do not round blindly to 39 and then call the exact statement true. With R = 39, 39/230 × 100% is approximately 16.9565%, which rounds to 17% to the nearest whole percent. With R = 40, 40/230 × 100% is approximately 17.3913%, which also rounds to 17%. Both counts are compatible with a displayed whole-number percentage of 17%. If the wording says precisely 17%, neither is exactly compatible with that whole number of 230.
The phrasing ‘17%’ and ‘17% to the nearest whole percent’ are not identical in a mathematical proof. Look for the precision statement, then decide whether an integer solution is required or an interval of possible counts should be returned.
A rounding interval is a set of possible truths
If a displayed percentage is 17% rounded to the nearest whole percentage point, the unrounded percentage lies from 16.5% inclusive to 17.5% exclusive under the usual rounding convention. For a total of 230, this gives 0.165 × 230 = 37.95 and 0.175 × 230 = 40.25. Valid whole-number R values are therefore 38, 39 or 40. That interval is wider than the two easy guesses made without setting out bounds.
This demonstrates why checking only the familiar decimal 0.17 can lose valid possibilities. The correct inequality is 0.165 ≤ R/230 < 0.175. Multiply through by the positive 230 and then select the integers satisfying it. Always account for the interval endpoints. Different rounding conventions at an exact half-way point should be stated if the question makes them relevant.
A good written conclusion is: ‘Assuming 17% was rounded to the nearest 1%, the true count could be 38, 39 or 40, so the exact count cannot be determined.’ Do not invent a single answer because the question seems to expect one.
Weighted percentages: combine the counts, not the headlines
Suppose Group A has 20 of 100 entries with the trait (20%), and Group B has 48 of 80 (60%). A simple average of 20% and 60% is 40%, but that would give equal weight to unequal groups. The combined total with the trait is 20 + 48 = 68, and the combined group size is 100 + 80 = 180. The combined fraction is 68/180 ≈ 37.8%, not 40%.
The general rule is to sum the numerators and sum the denominators, then divide. If group i has fraction p_i and size T_i, the combined proportion is (p_1 T_1 + p_2 T_2)/(T_1 + T_2), assuming consistent categories and no double counting. A weighted mean of the fractions with group sizes as weights produces the same result.
The arithmetic is straightforward; the interpretation is what matters. An equal-weight mean may be appropriate for a different question, such as the mean of two group percentages considered as two equally important observations. But if the question asks what percentage of all entries combined has the trait, weights are not optional.
A more deceptive weighted example
Class A’s fictional group has 10 of 20 entries with a trait, or 50%. Class B’s group has 5 of 100, or 5%. The unweighted average of percentages is (50 + 5)/2 = 27.5%. The combined proportion across all 120 entries is (10 + 5)/(20 + 100) = 15/120 = 12.5%. The two summaries answer different questions and differ by fifteen percentage points.
A poster that prints ‘average 27.5%’ without saying it is an average of group rates invites confusion. If the intent is to describe the proportion across all entries, 12.5% is the appropriate figure. Define the target before choosing a method; do not choose the formula just because it gives an impressive number.
That principle transfers to school attainment reports, transport surveys and sports participation statistics. When groups have different sizes, an overall rate is normally computed from totals. Keep your numerator and denominator attached to the population the claim is about.
Reverse a weighted average
Suppose a combined population has 400 entries and 35% carry a trait, so total R = 0.35 × 400 = 140. One subgroup contains 150 entries and 30 carry the trait. The remaining subgroup has 400 − 150 = 250 entries, with 140 − 30 = 110 carrying the trait. Its fraction is 110/250 = 44%.
The same work can be expressed as an equation. Let x be the number carrying the trait in Group B. Then (30 + x)/(150 + 250) = 0.35. Multiply by 400: 30 + x = 140. Therefore x = 110 and the required percentage is 44%. Using the single combined percentage as though it applied unchanged to each subgroup would be a serious error.
Always verify the reconstructed counts. Group A contributes 30, Group B 110, and their sum is 140. Group B’s count is nonnegative and no larger than its size of 250. Those checks show that the algebra gives a feasible whole-number solution.
The denominator can be hidden in words
A problem might refer to ‘all entries tested’, ‘all entries returned’, ‘valid entries’ and ‘entries with a readable result’. These can be different denominators. Suppose 200 records were collected, 20 were invalid and 45 of the 180 valid records carry a specified trait. Then 45/180 = 25% of valid records, while 45/200 = 22.5% of all collected records. Neither rate is inherently wrong, but each answers a different question.
If a question specifies ‘among valid records’, exclude invalid ones from the denominator. If it asks for a fraction of all collected records, include them. Do not count an invalid entry as negative or positive merely to avoid missing data. The validity rule comes from the stem, not from your preference.
State the denominator in your final sentence. ‘Twenty-five percent of valid records’ is informative; ‘twenty-five percent of records’ is ambiguous. Small wording choices prevent large errors.
Conditional percentages: within which group?
Imagine a table: in Group A, 18 of 60 entries carry a trait; in Group B, 36 of 90 do. Within Group A the rate is 18/60 = 30%. Within Group B it is 36/90 = 40%. The question ‘Of entries with the trait, what fraction came from Group A?’ uses a different denominator: 18/(18 + 36) = 18/54 = 1/3.
That reversal is common in data problems. ‘Percentage with the trait among Group A’ and ‘percentage from Group A among those with the trait’ change the condition. A quick two-way table makes it visible. Label rows by group and columns by trait presence; put totals at the margins. The phrase after ‘among’ or ‘given’ usually identifies the denominator.
English comprehension helps Mathematics here. Trace exactly which noun phrase the percentage modifies. If you read a condition incorrectly, even flawless division answers the wrong question.
Compare like with like across time
Suppose a report gives 30 of 150 at Time A (20%) and 45 of 300 at Time B (15%). The count rises from 30 to 45, a 50% increase; the recorded proportion falls by five percentage points. A student could accurately describe both. It would be wrong to label the later group as having a greater rate simply because 45 is larger than 30.
The same pattern appears when hospital visits, school attendance or survey responses are compared. More events can accompany a lower rate if the total at risk grows. The appropriate measure depends on the question. Sometimes the absolute workload matters; sometimes the proportion matters. Mathematics does not decide which is important without a defined objective.
When asked to assess a statement, rewrite it with the missing noun: ‘the number increased’ or ‘the rate decreased’. That simple repair forces the meaning into view.
Why a percentage can never prove a cause
Consider fictional proportions increasing from 10% to 20%. The calculation is exact, but it does not tell us why the pattern occurred. Mathematical treatment can show the size and direction of change; it cannot, by itself, establish the biological reason. If the problem asks about natural selection, the explanation belongs to a causal model supported by additional facts. A numerical pattern alone may also be compatible with different sampling methods or movements between groups.
This distinction does not reduce the value of Mathematics. It strengthens it by defining what the numerical evidence actually is. Mathematical questions often ask for a ‘comment’ or a ‘reasonableness check’. A good comment separates arithmetic certainty from scientific inference. You can write ‘the fraction increased’ confidently and ‘the cause is not determined by these records’ accurately.
Use Volume 0886 for the mechanism and this volume for the quantitative audit. Neither should substitute for the other.
Reconstruct two unknown subgroup sizes
Suppose a fictional programme has 120 entries in two groups. Group A contains twice as many entries as Group B. Let B = x and A = 2x, so 3x = 120 and x = 40. Thus Group A has 80 and Group B 40. If A has 25% with the trait and B has 50%, the counts are 20 and 20 respectively. The combined percentage is (20 + 20)/120 × 100% = 33⅓%.
The equal number with the trait in the two groups does not imply equal rates because the group totals are different. Group B’s rate is twice Group A’s rate. This example tests equations, fractions and verbal reading together. The algebra is short only because the quantities were defined correctly at the start.
Check that the fractions yield whole counts. Group A 25% of 80 is 20; Group B 50% of 40 is 20. The counts sum to 40, and 40/120 simplifies to one-third. The result is feasible.
Build a simultaneous-equation solution
Suppose Group A and Group B contain a total of 240 entries. Together they have 72 entries with a trait. Group A’s trait rate is exactly 20% and Group B’s exactly 40%. Let A = a and B = b. Then a + b = 240 and 0.20a + 0.40b = 72. From a = 240 − b, we obtain 48 − 0.20b + 0.40b = 72, so 0.20b = 24 and b = 120. Therefore a = 120.
The recorded counts are 0.20 × 120 = 24 in A and 0.40 × 120 = 48 in B. These sum to 72. Notice how a combined rate of 72/240 = 30% sits between the subgroup rates of 20% and 40%. That makes sense because both groups have the same size.
An algebraic result giving negative group sizes or more trait-bearing entries than the whole group would be infeasible and should trigger a check. The context adds constraints to the equations.
Use inequalities for decision thresholds
A fictional display may require that at most 25% of the included entries show a characteristic. If exactly 30 entries have the characteristic and the total is T, the requirement is 30/T ≤ 0.25. For positive T, multiply through: 30 ≤ 0.25T, hence T ≥ 120. The total must be at least 120 to satisfy this numerical threshold. That is a mathematical requirement, not a real policy recommendation.
If T must be a whole count, the smallest feasible T is 120. If T = 119, 30/119 is about 25.21% and fails the criterion. If T = 120, the rate is exactly 25% and passes if the rule uses ‘at most’. If the wording said ‘less than 25%’, equality would fail and the smallest integer total would be 121.
Read inequality words precisely: at least, more than, no more than and less than. A one-symbol difference can change the threshold answer.
Rounding at a boundary
Suppose a hypothetical quality report displays ‘25% with the trait’ rounded to the nearest whole percent, and the acceptance rule requires an actual fraction below 25%. The displayed figure alone does not settle the decision. The true fraction could be 24.6%, 24.9%, 25.0% or 25.4%, all of which display as 25% to the nearest whole percent under standard rules. Some would pass the strict threshold and others would not.
An examination answer should say that the rounded report is insufficient and request unrounded counts or greater precision. It should not infer that ‘25% displayed’ means exactly 25% measured. Likewise, when a bound from a measurement interval straddles the threshold, a guarantee is not justified.
The critical thinking is elementary but powerful: calculate the range of values compatible with the reported measurement, then see whether all of them satisfy the decision rule. If not, the conclusion must be conditional.
A feasible-range example with rounded counts
Suppose a set contains 180 entries and the report states that ‘about 30%’ have a trait, specifically a percentage rounded to the nearest whole percent. The true rate must lie from 29.5% inclusive to 30.5% exclusive. Multiply by 180 to obtain 53.1 ≤ R < 54.9. The only whole-number count satisfying that inequality is R = 54. Thus, in this case, the rounded percentage and total together determine the count uniquely.
Compare this with the earlier 17% of 230 case, which allowed three possible integer counts. Whether rounding creates ambiguity depends on the denominator and interval width. There is no universal rule that a rounded percentage always leaves several possible counts.
Show this working if the task asks for all feasible values. If it merely asks for the nearest estimate, the answer might be shorter; follow the command word and precision stated in the question.
Test a missing-data claim
A report says that in a group of 200 records, 60 show the trait, 20 are invalid and the remaining are recorded as without the trait. The 60 positive records are 30% of all collected records and 60/180 = 33⅓% of valid records. If someone claims ‘exactly one third of the entire group has the trait’, they have used the valid-record fraction as though it described every collected record.
The correct audit is to name the categories, sum them to 200 and choose the required denominator. There are 60 positive, 120 negative valid and 20 invalid. If the missing values were later resolved, the overall count with the trait could lie from 60 to 80, giving a whole-group percentage from 30% to 40%. That interval does not treat invalid entries as known positives; it describes the feasible extremes.
A bound is often more honest and more useful than pretending the missing records do not exist.
A two-period change with different sample sizes
Imagine Group A has 50 positive of 250 at Time 1, and Group B has 45 positive of 150 at Time 2. The rates are 20% and 30%, so the later rate is ten percentage points higher. The positive count decreases by five; its relative change is (45 − 50)/50 × 100% = −10%. The total decreases by 100; its relative change is −40%.
The rate comparison is valid as a comparison of recorded fractions. But if the groups do not sample the same underlying population, you should not call it the change in a single population without support. This is the bridge between mathematics and scientific inference: a correct numerical comparison does not erase a design limitation.
Use precise reporting language: ‘The later sample has a higher recorded proportion’ is safer than ‘the population became more resistant’ when the source has not established continuity.
A table that reverses the intuitive conclusion
Suppose a hypothetical combined report has two locations. In Location A at Time 1, 8 of 10 show the trait (80%), and in Location B at Time 1, 10 of 100 do (10%). The overall Time 1 rate is 18/110 ≈ 16.36%. At Time 2, Location A records 70 of 100 (70%), and B records 0 of 10 (0%). Each location’s rate has fallen, yet the combined Time 2 rate is 70/110 ≈ 63.64%, far higher.
This is possible because the mix of entries across locations changed dramatically: most entries shifted from the low-rate location to the high-rate location. The combined result is weighted by location sizes. This phenomenon is a valuable extension for advanced reasoning, not a requirement that every G3 student memorise a statistical name.
When an overall trend seems inconsistent with subgroup trends, inspect the group weights before alleging an error. The practical lesson is not ‘graphs lie’ but ‘the composition of the data can change the meaning of a combined average’.
Converting a long table into an efficient solution
For a multi-row question, make columns for group, total, number with trait, fraction, weight and any missing category. Fill each row before calculating overall percentages. If two rows have different units or definitions, do not combine them without correction. If the condition describes ‘new entries this week’ and another describes ‘cumulative entries since January’, adding the two double-counts overlapping records.
Write a sentence defining the target: ‘I need the percentage of all valid entries across both locations.’ Then use the valid totals. That sentence may feel unnecessary while practising, but it prevents many wrong-method errors in unfamiliar Paper 2 contexts. The examiner cannot award a correct numerical interpretation when the method answers a different question.
If the table is large, write partial sums with their category names. Numbers without labels quickly lose their meaning.
Exam practise 1: the shrinking denominator
Question: At A, 12 of 120 entries have a trait. At B, 10 of 40 do. Which count increased? Neither; the trait count fell by two. Which proportion increased? It rose from 10% to 25%, or by 15 percentage points. Did the count increase by 150%? No. The fraction increased by 150% relative to its starting fraction, but the absolute count decreased.
Model response: ‘Although the recorded number fell from 12 to 10, the rate rose from 10% to 25% because the total fell from 120 to 40.’ The explanation identifies the denominator rather than merely quoting the result. It is a short example of a longer skill.
Exam practise 2: reverse the total
Question: A report shows 84 entries with a characteristic, representing exactly 35% of all entries. What is the total? Let T denote the total. Set 84/T = 0.35, so T = 84/0.35 = 240. Then 240 − 84 = 156 entries lack the trait. Verify that 84/240 = 0.35. The result is a feasible integer total.
Common error: calculate 84 × 0.35 = 29.4. That operation finds 35% of 84, but 84 is already the specified part, not the whole. Mark part and whole before selecting a direction of calculation.
Exam practise 3: a weighted comparison
Question: Group P contains 30 of 50 with a trait; Group Q contains 14 of 70. Find the combined percentage. The counts are 30 + 14 = 44 positive of 50 + 70 = 120 total. The combined rate is 44/120 × 100% = 36⅔%, or about 36.7% to three significant figures. The subgroup rates are 60% and 20%; their unweighted mean is 40% and is not the combined percentage.
Model check: the combined rate should lie between 20% and 60%, and the larger 70-entry group with the lower rate should pull the combined rate below the simple midpoint. 36.7% passes both reasonableness checks.
Exam practise 4: calculate a missing subgroup
Question: There are 300 records, of which 90 have a trait. Of 100 records from Site A, 20 have the trait. What is the rate among the remaining records? The remaining total is 200 and the remaining positive count is 70. Therefore the remaining rate is 70/200 = 35%. The combined rate of 90/300 = 30% lies between Site A’s 20% and Site B’s 35%, as expected.
Do not subtract percentages (30% − 20% = 10%) to obtain the other group’s rate. The percentages are based on different denominators. Reconstruct counts first.
Exam practise 5: all possible integer answers
Question: A displayed rate is 12% to the nearest whole percent among 250 records. Determine all possible positive counts. Set 11.5% ≤ 100R/250 < 12.5%. Equivalently, 28.75 ≤ R < 31.25. Thus R may be 29, 30 or 31. Check: 29/250 = 11.6%, 30/250 = 12% and 31/250 = 12.4%, each rounds to 12%.
The point is not to guess one ‘likely’ count. The question asks for all values consistent with the given precision. Return the set with a short explanation of the rounding rule.
Exam practise 6: strict and inclusive bounds
Question: Thirty entries with a trait are already recorded. Find the smallest whole group total T so that the trait percentage is strictly below 20%. The condition is 30/T 150. The least integer T is 151. At T = 150 the rate is exactly 20%, which fails ‘strictly below’. At T = 151 the rate is about 19.87%, so it passes.
Had the rule said ‘at most 20%’, equality would have been allowed and the answer 150. Write the inequality before manipulating it, especially when the unknown is in the denominator.
Exam practise 7: two simultaneous rates
Question: Two groups contain 500 entries in total. Group A has 10% with the trait, Group B 30%, and 110 entries with the trait altogether. Let a + b = 500, 0.10a + 0.30b = 110. Substitute a = 500 − b: 50 − 0.10b + 0.30b = 110. Thus 0.20b = 60, b = 300, a = 200. The trait counts are 20 and 90, which sum to 110.
Check that both group totals are whole and nonnegative and that 110/500 = 22% lies between 10% and 30%. There are no impossible fractional entries in this example.
Exam practise 8: distinguish two directions of conditional probability
Question: In a fictional 200-record table, 40 of 80 records in Group A carry the trait, while 20 of 120 in Group B carry it. What percentage of Group A carry the trait? 40/80 = 50%. What percentage of all trait-bearing records belong to A? 40/(40 + 20) = 66⅔%. The same numerator has two different denominators because the questions are different.
Underline the condition after ‘of’. A calculation that uses an incorrect whole may look perfectly neat but fails to answer the request.
Exam practise 9: missing entries as a bound
Question: Among 150 records, 45 are positive, 90 are negative and 15 are invalid. What percentage of all 150 could be positive once invalid results are resolved? The minimum is 45/150 = 30% if none of the invalid entries is positive; the maximum is (45 + 15)/150 = 40% if all are positive. Thus the feasible overall percentage lies from 30% to 40%.
This interval is a possibility range, not a confidence interval or a probability statement. The data supply no basis to say how likely each value is.
Exam practise 10: check a public claim
Question: A notice says ‘Resistance doubled because the proportion rose from 25% to 50%.’ The proportion did double, but the wording ‘resistance doubled’ is ambiguous. It may be mistakenly read as the number of trait-bearing entries doubling. Without totals, the number cannot be inferred. A more accurate statement is: ‘The observed fraction was twice as high, changing from 25% to 50%; group sizes are not provided.’
Quantitative honesty often requires adding only five or six words. Those words specify the variable and keep the claim within the evidence.
Exam practise 11: total rates versus subgroup trends
Question: At Time 1, Site A has 8 of 10 positive and Site B 10 of 100 positive. At Time 2, Site A has 70 of 100 positive and Site B none of 10 positive. Verify that each site’s rate fell but the total increased. A fell from 80% to 70%. B fell from 10% to 0%. The overall rate changed from 18/110 ≈ 16.4% to 70/110 ≈ 63.6%.
The pattern is not an arithmetic contradiction. Most of the records at Time 2 came from the higher-rate site. A good answer names the changing group weights and avoids interpreting the combined rate without subgroup context.
Exam practise 12: construct a counterexample
Question: Show that an increasing proportion does not require an increasing count. Give an early total of 200 with 20 positive (10%) and a later total of 40 with 12 positive (30%). The positive count falls by eight, yet the fraction rises by twenty percentage points. The counterexample satisfies whole-number constraints and both totals are positive.
Writing a concrete valid counterexample is stronger than saying merely ‘not always’. It establishes exactly where an overgeneralised statement fails.
A decision checklist for longer Paper 2 problems
Before finalising, ask six questions in ordinary language. Which group is the whole? Do all rows use the same category definition? Am I comparing counts or proportions? If I combine groups, have I weighted by size? If the numbers are rounded, could more than one integer answer fit? Does the final claim require evidence about causes that the arithmetic cannot supply? Each question attacks a different likely failure.
When a task requires a recommendation, calculate first and recommend second. State any boundary: for example, ‘The numerical threshold is satisfied under the given definition and exact counts, but the data do not establish the biological mechanism.’ It is good Mathematics to know what a result means, and equally good Mathematics to know what it does not mean.
If you are short of time, protect the basic marks: write a clearly defined fraction, show substitution, put the unit or percentage sign, and make a reasonableness check. An extended comment can follow if the question awards it.
How to revise with a 20-minute rotation
Spend five minutes converting counts to fractions and vice versa. Spend five minutes solving one reverse total or missing subgroup problem. Spend five minutes interpreting a rounded percentage as an interval. Use the final five minutes to correct a deliberately misleading headline. Repeat across different contexts—school survey records, habitat observations, attendance data—so you practise the reasoning rather than memorise a single theme.
On the next session, attempt an unfamiliar mixed problem without notes. Read the entire stem once, mark every total and unit, and then decide the target. Review a wrong answer by locating the first changed denominator or unproved assumption rather than simply recording the correct final number. This kind of error analysis is much more valuable than doing twenty near-identical calculations without understanding why a mistake occurred.
Related foundational routes include PSLE Foundation Mathematics and subject-specific study methods.
How the four-volume sequence fits together
Volume 0886 explains why changing trait frequencies may be consistent with inherited variation and selection. This volume owns the counts, weighted percentages, reverse equations and bounds. The following English lesson addresses how a writer can make a responsible recommendation from supplied information without turning a rate into a medical promise. The EMS synthesis then asks whether the science, calculations and language agree.
The earlier Volume 0883 remains the owner for inhibition-zone circle geometry, not the present population-accounting problem. For a broader entry point to the exam system, see G3 SEC English, Mathematics and Science at Sengkang and the PSLE learning guide.
The central principle: the correct numerator, correct denominator and clearly defined comparison are more important than an impressive-looking percentage. If your answer is impossible as a count, inconsistent as a fraction, or broader than the source, repair it before moving on.
Official references and educational limits
Examination codes and the general SEC structure are verified against the SEAB 2027 G3 school-candidate syllabus index, updated 26 October 2026, and SEAB’s SEC overview. Recheck your own examination-year syllabus. The datasets here are invented, all rates and policies are classroom illustrations, and no problem should be used as clinical guidance.
Mathematics can tell us that a proportion rose, that a weighted average is wrong or that a rounded rate permits several values. It cannot, from these figures alone, tell anyone which treatment is appropriate. The value of the mathematics is precisely its discipline about what has been established.
