Six reports out of forty people do not mean the same percentage as six reports out of sixty people. The numbers are equal, but the denominators are different. In a fictional habitat survey, Group A has 40 participants and six symptom reports; Group B has 60 participants and six reports. The first proportion is 15%, the second is 10%, and the overall figure is 12%. All three can be correct at once. The examination skill is to know which population a percentage describes—and what conclusion it cannot support.
This G3 SEC Mathematics guide follows Vol 1250 on infectious-disease transmission and evidence limits. The previous Biology chapter established why a symptom report does not identify a pathogen or prove a waterborne cause. This chapter has a different owner intent: constructing and reading tables, choosing denominators, interpreting graphs, distinguishing percentages from percentage-point changes, and solving simple probability questions using clearly defined events. It also connects to Vol 1247 on batch rates, but does not repeat that volume’s production-capacity problem.
- Turn raw counts into a complete two-way frequency table.
- Choose between a group percentage, an overall percentage and a combined event.
- Calculate percentage points and relative percentage changes correctly.
- Interpret bar charts without being fooled by missing denominators or truncated axes.
- Use simple probability, complements, additions and tree diagrams only with justified assumptions.
- Distinguish sample comparisons from claims about causes and populations.
- Practise 36 original worked examples plus a complete Paper 2-style mini case.
What the official 2027 G3 Mathematics syllabus covers
The SEAB 2027 G3 school-candidate list, updated on 26 October 2026, identifies Mathematics as K310, with 4052 as the earlier O-Level reference code. The official K310 syllabus assigns Paper 1 and Paper 2 90 marks each, 50% each and 2 hours 15 minutes each. Paper 1 contains about 26 short-answer questions; Paper 2 contains nine or ten questions, and its last question focuses on real-world application. The syllabus includes data handling, statistical representations, interpretation of misleading diagrams and probability of simple and combined events.
Important exam habits include displaying essential working, keeping units and following required accuracy. The official syllabus states that non-exact numerical answers generally should be given to three significant figures, or one decimal place for angles in degrees, unless another precision is specified. In this guide, some exact fractions and percentages are left exact because the numbers divide neatly. Read the actual question’s instructions rather than rounding automatically. This invented survey is for mathematical literacy and exam preparation, not a prediction of a 2027 question or an epidemiological study.
Build the table before interpreting the story
The hypothetical records are: Group A has 40 participants, six of whom report a symptom; Group B has 60 participants, six of whom report one. The complete two-way frequency table can be described in text. Group A: six reported, 34 did not, total 40. Group B: six reported, 54 did not, total 60. Overall: 12 reported, 88 did not, total 100. Check every row and column: 6 + 34 = 40; 6 + 54 = 60; 6 + 6 = 12; 34 + 54 = 88; 40 + 60 = 100; and 12 + 88 = 100.
Students often start calculating from the story before setting out the structure. That makes it easy to divide by the wrong quantity. A labelled table is a compact picture of the information. It also creates a check: the total number of reported and unreported cases must equal the total number of participants. If the figures do not reconcile, stop. Find the missing or miscopied entry before proceeding to a formula. A wrong table can produce a page of perfectly executed but irrelevant calculations.
Choose the denominator that answers the question
“What percentage of Group A reported a symptom?” asks for 6 ÷ 40 × 100 = 15%. “What percentage of Group B reported one?” asks for 6 ÷ 60 × 100 = 10%. “What percentage of all participants reported one?” asks for 12 ÷ 100 × 100 = 12%. “What percentage of the entire survey were Group A participants who reported one?” asks for 6 ÷ 100 × 100 = 6%. These are distinct quantities despite using the same six people in several calculations.
Whenever you see the word “of”, point to the reference group. “Of Group A” means denominator 40. “Of all participants” means denominator 100. “Of people who reported symptoms” would mean denominator 12; half of those reports came from Group A because 6 ÷ 12 = 50%. A student who silently switches the reference group may make correct arithmetic but answer the wrong question. Write the denominator’s meaning beside the working at least once.
Percentage points are not a relative percentage increase
The Group A reported proportion is 15%; Group B is 10%. The absolute difference is five percentage points, calculated as 15 − 10. If comparing A with B as the starting reference, the relative increase is (15 − 10) ÷ 10 × 100 = 50%. Neither answer is universally right: “How many percentage points greater?” wants five; “What is the percentage increase relative to B?” wants 50%. The base matters. State both the number and the reference.
It would be misleading to say “there were 50% more symptom reports in A”. Both groups have six reports, so the count was not greater. The rate was 50% higher relative to B under this simple comparison. The group sizes differ, which is exactly why counts and proportions must be separated. The Mathematics answer describes numerical relationships; whether the different reported proportions reflect any underlying cause remains outside what the table alone can prove.
From table to graph without losing meaning
A bar chart of counts would give both groups a bar height of six. Such a chart answers “How many reports were recorded?” It does not on its own show the different group-specific proportions. A bar chart of percentages would show 15% for A and 10% for B and answer “What proportion of each group reported?” Neither chart is dishonest if correctly labelled. The problem begins when a speaker uses a count chart to claim equal rates, or a rate chart to claim that the raw number of reports in A was larger.
Axes should start at a scale that lets the audience judge differences fairly. An axis cropped from 9% to 16% can make a five-percentage-point gap look enormous. A zero-based percentage bar chart communicates the scale more proportionately, while a clearly labelled non-zero scale may still be acceptable in contexts where the purpose is to inspect small variation. The key examination habit is to explain how the chosen scale affects the reader’s interpretation rather than declaring every non-zero axis invalid.
Mean, median and range are not interchangeable descriptions
Suppose a separate fictional dataset records symptom-report counts across five survey days: 0, 1, 1, 2, 8. The total is 12, so the mean is 12 ÷ 5 = 2.4 reports per day. The ordered median is 1, the middle entry; the range is 8 − 0 = 8. These descriptions are all correct. The high value of eight has pulled the mean above the typical middle value. A learner should explain which quantity is useful for the stated task rather than claiming that the mean is “wrong”.
Compare a second series, 2, 2, 2, 3, 3. Its total is also 12 and its mean is also 2.4. Yet its range is 1 rather than 8. Equal means can hide different spreads. The syllabus includes measures of spread such as range, interquartile range and standard deviation. This short comparison is a foundation; for a formal spread question, select the measure named or requested and show any necessary calculation. Do not confuse a spread in daily report counts with proof of a biological cause.
Probability begins by naming the random experiment
Imagine one participant is selected uniformly at random from the 100 hypothetical survey entries. The probability that the selected entry belongs to Group A is 40/100 = 0.4. The probability that it is a symptom-report entry is 12/100 = 0.12. The probability that it is both Group A and a symptom-report entry is 6/100 = 0.06. If the event is “Group A or reported a symptom”, count 40 + 12 − 6 = 46 entries; the probability is 46/100 = 0.46. The subtraction prevents double-counting the six entries in both groups.
Probability is not meaningful without the sampling rule. Uniform selection from these 100 entries creates one finite chance experiment. It does not tell us a person’s actual chance of becoming ill from an environmental exposure. If the records came from an unrepresentative sample, selecting a random record does not automatically represent a random person in the full habitat population. The mathematical fraction can be exact for the specified list while its generalisation remains unjustified.
Independent events must be an assumption, not a convenience
Suppose two entries are selected in succession with replacement from the 100-entry list, and each time selection is uniform. The probability that both selected entries are symptom reports is 12/100 × 12/100 = 0.0144, or 1.44%. Replacement restores the original list, so the two selections are independent under the stated setup. This is a probability about selecting record cards, not about the chances that two people contract a disease.
Without replacement, the second selection is from 99 remaining entries. If the first was a symptom-report entry, 11 such entries remain; the probability of two reports is (12/100) × (11/99) = 132/9900 = 1/75, approximately 1.33%. The difference is small here, but the method matters. Never use an independent-events formula merely because it is shorter. Draw a tree diagram, identify whether the composition changes, and label the branches with denominators that match the experiment.
Check the complement when it makes the calculation easier
The probability that a uniformly selected entry from the 100 records has no symptom report is 88/100 = 0.88. It is also 1 − 0.12 = 0.88. Complements are especially useful in combined-event questions, but the event must be precisely defined. The complement of “reported a symptom in the survey” is “did not report a symptom in the survey”. It is not “was certainly healthy” or “had no exposure”, because those are different claims not documented in the dataset.
Why survey mathematics does not settle Biology
Mathematics describes the data actually given. Biology investigates biological processes and the strength of evidence connecting them to outcomes. The values 15%, 10% and 12% are exact for the invented survey. They do not identify a pathogen, establish that recovered water was consumed, or prove transmission through an outlet. The previous Biology volume holds those explanatory boundaries. When a Mathematics real-world question includes a health context, a responsible conclusion may need both a correct numerical comparison and an explicit limit on what the comparison implies.
Take the sentence “Group A had a higher reported rate, so its water caused the symptoms.” The first clause follows from 15% versus 10%. The second is not supported by the table. The corrected sentence is: “The reported proportion was five percentage points higher in Group A than in Group B in this sample; the data alone do not establish a cause.” That is the difference between interpreting the values and inventing a mechanism.
Thirty-six worked numerical and interpretation clinics
Clinic 1: What is the Group A non-report count?
Subtract the six reports from the Group A total: 40 − 6 = 34. Check the row by adding 34 and six to recover 40. A two-way table is complete only when both categories and their totals reconcile.
Clinic 2: What is the Group B non-report percentage?
Fifty-four of 60 participants did not report the symptom. The percentage is 54/60 × 100 = 90%. This is the complement of the Group B report percentage of 10%, but it is not proof that 90% were medically healthy.
Clinic 3: What fraction of all participants belong to Group B?
Group B has 60 of the 100 entries, so the fraction is 60/100 = 3/5. Do not use 6/60; that fraction answers the different question of symptom-report proportion within Group B.
Clinic 4: What is the overall report percentage?
Add reports first: 6 + 6 = 12. Add all participants: 40 + 60 = 100. Hence 12/100 × 100 = 12%. Taking the unweighted average of 15% and 10% gives 12.5%, which is wrong for unequal group sizes.
Clinic 5: How many percentage points separate the report proportions?
The values are 15% and 10%, so the gap is 15 − 10 = five percentage points. In precise prose, “five percentage points” is not the same as “five per cent more” relative to a reference.
Clinic 6: What is A’s relative percentage excess compared with B?
Use B as the base: (15 − 10)/10 × 100 = 50%. The rate in A is 50% higher relative to B. This says nothing about a greater number of reports, since both groups recorded six.
Clinic 7: What fraction of all reports came from A?
There were 12 reports in total, six from A, so 6/12 = 1/2 = 50%. The denominator is the set of reporters. A learner who divides by 40 obtains the within-A reporting rate instead.
Clinic 8: What is the probability of choosing an A participant?
Under uniform random selection of one of the 100 records, the probability is 40/100 = 0.4. The result depends on the defined selection of entries; it is not automatically a probability for every person in a wider population.
Clinic 9: What is the probability of selecting a non-report entry?
There are 88 records without a report. Probability = 88/100 = 0.88. You can also use the complement 1 − 12/100 = 0.88. Name the event precisely: non-report, not confirmed absence of illness.
Clinic 10: What is the probability of A and a report?
The intersection contains six entries out of 100, so the probability is 0.06. Do not multiply 0.4 by 0.12 unless independence has been established; the table itself gives the joint count.
Clinic 11: What is the probability of A or a report?
Count the union: 40 A records + 12 reports − 6 records counted twice = 46. Thus 46/100 = 0.46. The subtraction is the important step; otherwise overlapping entries are counted twice.
Clinic 12: What is the probability of B and no report?
Group B contains 54 non-report entries. The probability under uniform selection from all records is 54/100 = 0.54. If the question instead says “from Group B”, the answer is 54/60 = 0.9.
Clinic 13: How should the chart label its vertical axis?
If comparing group reporting rates, write “Participants reporting a symptom (%)”, not merely “Cases” or “Water quality”. Axis labels should reveal the measured quantity and denominator basis. A correct bar height with an incorrect label is misleading.
Clinic 14: Why can a count chart show two equal bars?
Both groups have six reports, so a count chart legitimately gives equal heights. The groups differ in size, so equal counts do not imply equal percentages. Select or describe the representation that fits the question.
Clinic 15: What does a zero horizontal axis show?
A bar chart beginning at zero lets readers compare proportions against the full zero-to-value span. A cropped scale can magnify perceived differences. In evaluation, explain the effect rather than mechanically rejecting every chart with a non-zero axis.
Clinic 16: Is the difference between 15% and 10% a 50-percentage-point gap?
No. The absolute gap is five percentage points. The figure 50% is a relative increase using 10% as the reference. Use the named comparison as the guide to the correct operation.
Clinic 17: Why is the average of 15% and 10% not the overall percentage?
The groups contain 40 and 60 people, so their proportions need weighting by the group sizes. Calculating from counts gives (6 + 6)/(40 + 60) × 100 = 12%, not 12.5%. The equal-weight mean assumes equal-sized groups.
Clinic 18: What is the probability of two reports with replacement?
Each uniform selection has probability 12/100 and the list is restored. Multiply: (12/100)² = 0.0144. The assumption of independent repeated selections must be stated before multiplying.
Clinic 19: What is the probability of two reports without replacement?
After selecting one report, 11 reports remain among 99 records. Multiply 12/100 × 11/99 = 1/75, about 0.0133. The changed denominator distinguishes this from the replacement case.
Clinic 20: What is the probability of two non-reports with replacement?
With replacement, 88/100 is the probability on each uniform draw. Thus (0.88)² = 0.7744. This is a card-selection exercise; do not turn it into a probability of health outcomes.
Clinic 21: What is the probability of at least one report in two replacement draws?
Use the complement. Probability of no reports on either draw is 0.88² = 0.7744. Therefore the chance of at least one report is 1 − 0.7744 = 0.2256, or 22.56%.
Clinic 22: What is the probability that a randomly selected report comes from B?
Within the twelve report entries, six are from B, so 6/12 = 50%. The question conditions the choice by saying a report is already selected; the relevant denominator is twelve, not 100.
Clinic 23: What is the probability of no reports within A?
If a record is selected uniformly from Group A only, 34/40 = 0.85. State the selection pool explicitly. If the selection is from the entire survey, the intersection event “A and no report” has probability 34/100.
Clinic 24: What is the mean of 0, 1, 1, 2, 8?
The total is twelve across five observations, so the mean is 12/5 = 2.4. This is a mean of daily counts, not a percentage of participants. Keep the type of quantity visible.
Clinic 25: What is the median of 0, 1, 1, 2, 8?
The values are already in ascending order. The third of five observations is 1, so the median is 1. Unlike the mean, the middle value is less influenced by the large count of eight.
Clinic 26: What is the range of 0, 1, 1, 2, 8?
Range = largest value minus smallest value = 8 − 0 = 8. It describes the spread of observed daily counts without indicating why the days differed.
Clinic 27: Can two series have the same mean but different ranges?
Yes. Both 0,1,1,2,8 and 2,2,2,3,3 total twelve across five days, so each has mean 2.4. Their ranges are eight and one respectively. This shows why central tendency and spread answer different questions.
Clinic 28: What is the complement of ‘reported a symptom’?
Within the survey coding, the complement is “did not report a symptom”. The total is 88 out of 100. It is not logically the same as “did not experience any symptom” if reporting was incomplete.
Clinic 29: How can a diagram hide the different denominators?
A pie chart of twelve reports alone gives six and six equal halves. It does not show that the groups had 40 and 60 participants. To compare report rates, provide the group totals or use a clearly labelled percentage chart.
Clinic 30: Why is ‘50% more symptoms in A’ incorrect?
The count of symptom reports in A is six, equal to B’s six. What differs is the proportion within each group. Say “A’s reported proportion is 50% higher relative to B’s 10% rate” if that is the comparison intended.
Clinic 31: What if A had eight reports out of 40?
The new within-A proportion would be 8/40 × 100 = 20%. If B stays at six of 60, the overall proportion becomes 14/100 = 14%. Changing one cell also changes the row complement and column totals, so update the whole table.
Clinic 32: What if both groups had equal sizes of 50?
If each had six reports, each rate would be 6/50 = 12%, and the overall rate would also be 12%. Equal-sized groups remove this particular weighting issue, but they still do not establish any cause of reported symptoms.
Clinic 33: What if the target question asks about ‘at least one’?
Use the complement when it reduces steps. For two independent replacement draws, “at least one report” is one minus the probability of zero reports. A tree diagram can confirm that the event includes one report on either draw or two reports.
Clinic 34: When is a probability model not a real-world risk estimate?
A random-card experiment on 100 recorded entries has clear mathematical outcomes. It does not automatically represent the wider habitat population or a future health event. Probabilities depend on how the sample and random process are defined.
Clinic 35: Why must the table’s totals be checked first?
If the group totals or status totals do not reconcile, subsequent percentages could be built on faulty counts. Verify row and column sums before any fraction. In Paper 2, this small check can prevent a chain of costly errors.
Clinic 36: What is the safest conclusion after comparing 15% and 10%?
State that the reported proportion was five percentage points higher in Group A in this invented sample. Then identify that the counts alone do not show cause. This conclusion is numerically complete and proportionate to the evidence.
One longer Paper 2-style real-world problem
Part A. An imaginary habitat committee receives 150 survey entries. Group C has 50 people, of whom ten report a symptom; Group D has 100 people, of whom twelve report one. Complete the table. Solution: C has ten reports and 40 non-reports; D has twelve reports and 88 non-reports. Overall there are 22 reports, 128 non-reports and 150 entries. The totals check: 10 + 12 = 22 and 40 + 88 = 128. Start with these facts before writing percentages.
Part B. Calculate the reporting proportions and compare. Solution: C = 10/50 × 100 = 20%; D = 12/100 × 100 = 12%; overall = 22/150 × 100 ≈ 14.7% to three significant figures. The gap between C and D is eight percentage points. Relative to D’s rate, C’s is (20 − 12)/12 × 100 ≈ 66.7% higher. These are two different comparisons, not interchangeable descriptions.
Part C. A poster says “Most symptom reports came from the higher-rate group.” Is this accurate? Solution: No. The higher-rate group is C with 20%, but it has ten reports while D has twelve. A majority of all 22 reports came from D, since 12/22 ≈ 54.5%. The poster confuses a within-group proportion with the number of reports contributed to the total. The appropriate correction is to state which comparison is being made.
Part D. A random entry is drawn uniformly from the 150 records. Find the probability it is from C or is a symptom-report entry. Solution: There are 50 C entries and 22 report entries, with ten in both sets. The union contains 50 + 22 − 10 = 62 records. Probability = 62/150 = 31/75 ≈ 0.413. State the finite record-selection model; do not interpret this as a future illness probability.
Part E. The poster claims that the outlet used by Group C caused an infection. Evaluate the conclusion. Solution: The mathematical comparison finds a higher reported proportion in C, but the table lacks confirmed diagnostic and exposure information and does not isolate a causal factor. It supports a difference in this survey, not a conclusion about a pathogen or outlet. A final sentence that states the limit shows careful interpretation rather than a failure to answer the question.
How to mark your own work under examination pressure
When you check a question, begin with meaning, not arithmetic. Circle the population described by the denominator. Verify each row and column total. Check whether the result is an exact count, a percentage, a percentage-point difference or a probability. Read the event definition again before using union, intersection or complement rules. For repeated selections, identify whether the process includes replacement or explicitly justified independence. Then check the units, rounding requirement and final interpretation.
In an extended question, a result may be numerically correct but narratively wrong. Six reports in both groups is correct; “equal reporting rates” is not. A rate difference of five percentage points is correct; “50 percentage points” is not. A sample probability of 0.12 is correct for the defined record-selection experiment; “a resident has a 12% risk of infection” is not. The final sentence should preserve the question’s meaning and the bounds of the data.
A seven-day route from foundation to advanced comparison
Day one: rebuild two-way tables from word problems. Day two: practise group, overall and selected-subgroup denominators with new figures. Day three: distinguish percentages from percentage-point changes and relative changes. Day four: compare counts, rates and visual representations, explaining at least one misleading chart. Day five: solve single-event, union, intersection and complement probability tasks. Day six: use simple trees to distinguish replacement and no-replacement models. Day seven: complete a mixed Paper 2 problem, mark your first unsupported interpretation and repeat it with changed numbers.
Keep an exact error log. “I averaged percentages without weighting,” “I used the wrong group total,” “I double-counted the intersection,” “I forgot replacement,” and “I confused association with cause” are repairable entries. “I am bad at probability” is not a useful diagnosis. Parents helping a student can ask one precise question: “What exactly does the denominator represent?” If the learner can answer that confidently, much of the problem structure becomes clearer.
Frequently asked questions
Will a water-exposure survey appear in the 2027 examination? No prediction is made. This is original context-based practice. Is the 12% overall rate an average of 15% and 10%? No; the groups are different sizes, so use the total counts. Does a difference in rates prove a biological cause? No. Must I learn advanced medical statistics? No; the exercises focus on the K310 data and probability skills described in the official syllabus. Can I use a calculator? Follow the official paper instructions and approved-device rules; always show essential working.
Continue the learner route
The canonical Biology explanation for disease, agents and transmission is Vol 1250. For a different Mathematics intent, Vol 1247 explains batch recovery, rates and feasible schedules, while Vol 1243 explains time-weighted measurement in an air-quality context. For foundational routes, see G3 Mathematics foundations, PSLE Mathematics: The Voyage of Water and G3 preparation after PSLE. The next English and EMS guides can apply these newly learned numerical distinctions to careful public explanations. Always consult the official 2027 K310 syllabus for assessed content and requirements.
