A two-way table is a small accounting system. Each participant belongs to exactly one row and one column when the categories are mutually exclusive and cover everyone in the survey. Missing entries are not guesses: they can be recovered from row totals, column totals and a grand total. The strongest G3 Mathematics answers show where each number came from and confirm that all margins balance.
This learner’s guide uses a fictional school activity-and-refreshment survey to teach a transferable SEC Mathematics skill. It does not assess anybody’s heart health or prove that a habit causes a medical outcome. The 2027 G3 SEC Mathematics syllabus is K310, formerly 4052. Its statistics and everyday problem-solving objectives make careful reading of data, percentage bases, interpretation and checking essential. The 2027 papers are two compulsory 90-mark, 2-hour-15-minute papers, equally weighted. We will build the table first, then handle multi-step questions without changing denominators midway.
- Identify the population and how each record was classified before performing arithmetic.
- Place the grand total, row totals and column totals around the interior cells.
- Use subtraction to recover missing cells, and check both directions independently.
- Read ‘of the group’ as a denominator instruction; conditional percentages use the named group’s size.
- Keep counts, percentages and percentage-point comparisons separate.
- Use data language: classroom survey association is not a clinical diagnosis or proof of causality.
Where this chapter belongs in the series
The preceding Volume 0577 on coronary heart disease taught the biological mechanism and the limits of lifestyle claims. Its examples need numerical discipline when a poster cites percentages. Volume 0576 owns pulse time-series gradients. This chapter does not repeat those calculations. It owns a particular examination operation: rebuilding a missing two-way table and choosing the correct base for each question.
Learners transitioning from primary word problems can use the PSLE Learning Guide for totals and fractions, then the Mathematics Hub for secondary percentages, algebra and statistical thinking. This example is a teaching dataset, not a report about students at any school.
Check the examination year and code
The official 2027 K310 syllabus has Paper 1: 90 marks in 2 hours 15 minutes, with roughly 26 short-answer questions. Paper 2 also has 90 marks in 2 hours 15 minutes, with nine or ten longer questions and a final extended real-world application. Both papers are compulsory, each contributing 50%. Candidates should show working clearly. Unless a problem specifies otherwise, non-exact numerical answers are generally given to three significant figures, while angles are treated separately as the syllabus instructs.
The precise subject details can be checked at SEAB’s 2027 G3 syllabuses and the K310 Mathematics syllabus PDF. A real-world data question is not permission to invent a survey result. It is a chance to use ordinary arithmetic and reasoning correctly in an unfamiliar setting.
The original invented survey: know the people and the two questions
Imagine 120 volunteers completing a school learning survey. Each gives exactly one answer to each of two questions: did they meet a stated activity-plan target during the survey week (yes or no), and did they choose water at a specified refreshment break (yes or no)? The categories are deliberately simplified for a mathematics exercise. The questions concern reported behaviour, not clinical disease, and the sample is not presented as representative of Singapore.
The problem supplies four facts: 120 valid records altogether; 72 respondents met the activity-plan target; 58 chose water; and 40 both met the target and chose water. Every number has a place in the table. The remaining three interior cells are unknown. Your task is to reconstruct them without inventing a new person or counting one respondent twice.
Build a two-way table from the inside out
Write the row labels ‘target met’ and ‘target not met’. Write column labels ‘water chosen’ and ‘water not chosen’. Place the grand total 120 at the bottom-right intersection of total row and total column. Enter 72 as the first row total, 58 as the first column total, and 40 in the ‘target met and water chosen’ cell. The rest must satisfy these known margins.
Within the 72 who met the activity target, 40 chose water, leaving 72 − 40 = 32 who did not. Of the 58 who chose water, 40 met the target, leaving 58 − 40 = 18 who did not. The second row total is 120 − 72 = 48, so the last interior cell is 48 − 18 = 30. The second column total is also 120 − 58 = 62, and 32 + 30 = 62 confirms the other margin.
The finished table, stated in words
Target met: 40 chose water, 32 did not, total 72. Target not met: 18 chose water, 30 did not, total 48. Water chosen column: 58 overall. Water not chosen column: 62 overall. Grand total: 120. The four interior counts sum to 40 + 32 + 18 + 30 = 120. The two row totals sum to 72 + 48 = 120. The two column totals sum to 58 + 62 = 120.
If any of those three totals differs, stop before answering the percentage question. A wrong count can sometimes produce a plausible-looking percentage. Checking both margins is the fastest way to catch the initial error. Do not mistake the fact that row totals and column totals each sum to 120 as an instruction to add them together: they are two separate descriptions of the same 120 respondents.
Why the four interior cells should be mutually exclusive
Each participant must fall in one and only one interior cell. The phrase ‘both met the target and chose water’ identifies an intersection, not an extra fifth category. When an exam question describes ‘at least one’ category, overlapping membership may require a different structure. In our table, the two questions yield four distinct combinations, and all valid respondents are recorded once.
An important practical limit follows. If the original survey allowed multiple refreshment choices, or someone answered ‘sometimes’ to activity, the four-cell design would no longer describe all responses without a clearly stated classification rule. Mathematics begins with correct modelling. Recovering entries from an unsuitable table is not a victory even if the arithmetic balances.
Read conditional language before the calculator
Consider ‘What percentage of those who met the activity target chose water?’ The population named by ‘those who met the activity target’ is 72, and the matching subset is 40. Calculate 40 ÷ 72 × 100% = 55.6% to three significant figures. The denominator is not 120, because the question asks about a particular row rather than all respondents.
Now ask ‘What percentage of water choosers met the activity target?’ The population has changed to the 58 who chose water. The numerator happens to remain 40, but the denominator becomes 58. Calculate 40 ÷ 58 × 100% = 69.0% to three significant figures. Both answers can be correct. The sentences ask two different questions, even though they refer to the same intersection cell.
A third percentage: the whole sample
To find the percentage of all valid respondents who both met the activity target and chose water, use 40 ÷ 120 × 100% = 33.3%. This is neither 55.6% nor 69.0%. The word ‘all’ establishes the grand total as the base. The three distinct answers demonstrate why learning a formula without reading the sentence is not enough.
Make a habit of writing a small verbal fraction first: ’40 out of 72′, ’40 out of 58′, or ’40 out of 120′. If you cannot name the base, postpone calculator work. When a multiple-choice option looks familiar but belongs to another denominator, that familiarity is exactly the distractor the question is testing.
Two rows, two percentages: compare like with like
Among the 72 respondents who met the activity target, 40 chose water: about 55.6%. Among the 48 who did not meet the target, 18 chose water: 37.5%. The difference between these row-conditional percentages is approximately 18.1 percentage points. This comparison holds the selected behaviour (‘chose water’) constant while using each group’s own total.
A count-only comparison, 40 against 18, is less informative about rates because the groups are unequal in size. A percentage comparison helps answer how common that choice was within each group. Nevertheless, the percentages describe this invented sample. They do not show that one behaviour caused the other, that any participant was healthier, or that a public-health programme succeeded.
Percentage points and relative change are not the same
The numbers 55.6% and 37.5% differ by 18.1 percentage points when using the displayed rounded rates. If a question instead asks for the relative increase from 37.5% to the unrounded 55.555…%, calculate (55.555… − 37.5) ÷ 37.5 × 100%, approximately 48.1%. These are different quantities. State the base explicitly rather than saying ‘the first group is 18% higher’ without identifying what that means.
The general concepts of percentage-point changes appear elsewhere in the series, notably Volume 0383. Here the new challenge is to reconstruct the four cells and choose the conditional base before any comparison. Link these two skills, but do not treat this chapter as a replacement for the earlier general owner.
Reverse a percentage to obtain a count
Suppose a different group contains 80 people and exactly 35% chose a stated option. Multiply 80 by 0.35 to recover 28 people. The result is an integer, as a count must be. If the percentage was rounded to one decimal place, however, an exact reverse count may not be uniquely determined. Read whether the problem states ‘exactly’, ‘approximately’, or ’rounded to the nearest percent’.
Imagine 41% of 82 respondents supposedly selected an option. Multiplication gives 33.62 people, an impossible literal count. This tells you either the percentage has been rounded or the statement has been simplified. Do not silently write 33.62 people into a frequency table. Look for the original count or the specified rounding method.
Unknown totals: use simple algebra instead of random trials
A table may give a row total of x people and tell you that 25 of them fall in the first column. The second cell is x − 25. If the next row totals 90 − x and contributes 20 to the first column, the first column total is 45. A further equation about the second column can determine x. Algebra works because every total imposes a constraint.
Write the relationship before solving. For a two-by-two table with interior entries a, b, c and d, row totals are a + b and c + d; column totals are a + c and b + d; grand total is a + b + c + d. These identities are not advanced statistics. They are careful counting. They let you solve unfamiliar tables even when the story changes.
Redundant information should agree, not confuse you
A well-written question may provide both a grand total and several row or column totals. Some facts will be mathematically redundant. Use them as a check. If 72 met the target and 48 did not, the supplied grand total should be 120. If it says 125, something in the data or interpretation needs attention. Do not force the four-cell table to satisfy contradictory numbers by quietly altering a figure.
In a real dataset, different totals can arise from missing responses. That would require an explicit valid-record rule: for example, 125 submitted forms but only 120 with both questions completed. The table should make that denominator visible. In an exam, read every qualifying phrase about non-responses before assuming that all forms and all observations are counted.
Rounding must not create people
Counts are whole numbers. Percentages may be rounded for display. When a table says 28 of 80, preserve that exact count and compute 35% if necessary; do not reconstruct 28 from a displayed rounded percent when the original is already available. When comparing close percentages, use unrounded intermediate fractions and round only your final answer to the requested precision.
SEAB’s K310 numerical-answer instruction encourages appropriate three-significant-figure reporting unless otherwise stated. An exact frequency of 28 people is not ‘28.0 people’ merely to create three significant figures. An exact count remains an integer. Explain the difference between exact values and calculated approximations whenever the precision of the answer matters.
How to describe a two-way table in words
A strong description identifies the sample, the variable being compared, the group totals and the relevant within-group proportions. For example: ‘In this hypothetical sample, 40 of 72 respondents who met the activity target chose water (55.6%), compared with 18 of 48 respondents who did not meet the target (37.5%).’ The sentence contains enough information for a reader to check the comparison.
Do not shorten it to ‘active students drink more water’, which can imply amount or frequency rather than one recorded choice at a specified break. Nor should you call the two survey variables medical outcomes. Mathematical clarity is partly verbal precision. If a chart title or caption is ambiguous, rewrite it to specify what was actually measured.
A proportion is not a probability model for every population
You may describe 40/72 as the observed proportion in the activity-target group. A probability question might use those numbers as empirical estimates under specific assumptions, but the dataset alone does not establish the population probability for every learner. Sampling design, missing records and group definitions matter. Do not extrapolate from a fictional teaching survey to the wider public.
If a problem explicitly asks for the probability that one uniformly randomly selected valid respondent is in a particular cell, the probability is that cell count divided by 120. The uniform random selection rule is essential. A person selected only from the ‘target met’ row has a different sample space, and therefore a different denominator. The procedure must agree with the wording.
Three distinct kinds of missing data
A blank in a mathematical puzzle may represent an unknown but recoverable count. A blank in a real survey may represent a participant who did not answer. A blank in a printed chart could be an omitted label. These meanings are not interchangeable. Only the first case lets you reconstruct an exact number from sufficient margins under a complete-classification assumption.
If five people did not answer the refreshment question, decide whether they are outside the two-way table or whether a separate ‘no response’ column is required. You cannot allocate them proportionally just to make your preferred comparison look tidy. Validity comes from the described data collection, not from aesthetic symmetry.
From primary totals to secondary interpretation
At a primary level, you might be asked how many people chose water if 58 did, or how many did not if there were 120. At G3, the same subtraction sits inside a larger interpretation task. You must reconstruct missing cells, choose row or column bases, compare rates, check the constraints and explain why a headline exceeds the evidence. The arithmetic is familiar; the reasoning is richer.
Use the English Hub when the hard part is an ambiguous sentence such as ‘40% more chose water’. Use the Science Hub to understand what a biological mechanism can establish. This article remains the owner of the arithmetic problem: recovering counts in a two-way table and determining the relevant percentage base.
Timed-exam working: write four things in the right order
Start with the table labels. Enter the totals. Recover interior entries using the simplest row or column subtraction. Check the grand total twice. Only then write each requested ratio, keeping its words beside the figures. This takes slightly more space than doing calculations in your head, but it prevents a common loss of marks from an unexplained, incorrect denominator.
If time is running short, a fully labelled small table can make your method visible even before the final comparison sentence is complete. When a question asks for an explanation, write one sentence identifying the comparison base and one describing the evidence. Do not replace the explanation with a naked calculator display.
Exam practice clinic: thirty-five original mini problems
Case 01 — One missing row cell
Question: There are 65 respondents in a row; 29 selected option A. How many selected the only alternative?
Worked answer: Subtract within the same row: 65 − 29 = 36. State that the two options are mutually exclusive and cover every valid response in that row. The answer is a count of people, not a percentage; adding other rows would change the meaning and must not be done.
Case 02 — One missing column cell
Question: The column total for water choices is 44 and 27 of these came from Group A. How many came from Group B?
Worked answer: The second cell in that column is 44 − 27 = 17. Both groups must be the only categories contributing to the column. A column total is a count of one option across groups; it is not the denominator for every percentage in the table.
Case 03 — A missing row total
Question: A survey has 150 complete records and 86 fall in the first of two exhaustive rows. Find the second row total.
Worked answer: Subtract from the grand total: 150 − 86 = 64. Check that 86 + 64 = 150. Do not assume that each row must contain 75 people. A two-way table is a record of actual categories, not an automatically equal allocation.
Case 04 — The last interior cell
Question: In a 100-record table, row totals are 55 and 45. In the second row, 19 selected the first column. Find the other cell.
Worked answer: It is 45 − 19 = 26. This uses the second row total directly; no other value is needed. To check, add 19 + 26 = 45. If you also know a column total, use that separate constraint to verify the cell.
Case 05 — Recover all four cells
Question: A 90-person table has row A total 54, column X total 47, and intersection A-and-X 30. Rebuild the four interior cells.
Worked answer: A-X = 30; A-not-X = 54 − 30 = 24; not-A-X = 47 − 30 = 17; not-A-not-X = 90 − (30 + 24 + 17) = 19. Row totals are 54 and 36; columns 47 and 43. Check both margins independently.
Case 06 — Row conditional percentage
Question: Among 54 respondents in group A, 30 chose X. What percentage of group A chose X?
Worked answer: Use 30/54 × 100% = 55.6% to three significant figures. The words ‘of group A’ define the denominator. Dividing by the grand total 90 would answer a different question. Show the numerator and denominator before rounding.
Case 07 — Column conditional percentage
Question: Among 47 respondents who chose X, 30 are in Group A. What percentage of the X column is Group A?
Worked answer: Use 30/47 × 100% = 63.8% to three significant figures. The intersection count of 30 is the same as in the previous case, but the base is the entire X column, 47. This is why reversing the wording changes the answer.
Case 08 — Overall intersection percentage
Question: How much of the entire 90-person sample belongs simultaneously to Group A and chose X, if that intersection contains 30?
Worked answer: Calculate 30/90 × 100% = 33.3%. The word ‘entire’ identifies the grand total. Avoid choosing a denominator because it sits nearest the intersection cell. The question asks about all respondents and the exact fraction is one third.
Case 09 — A frequency and a percentage
Question: A poster writes ‘30% chose X’ although the table has 30 X selections out of 90 participants.
Worked answer: Thirty is a count, not 30%. The actual overall fraction is 30/90, approximately 33.3%. A number printed beside a bar does not become a percentage until it has been divided by a relevant total and multiplied by 100.
Case 10 — Unequal group sizes
Question: Group A has 30 of 54 choosing X; Group B has 17 of 36 choosing X. Which group has the higher within-group proportion?
Worked answer: Group A: 30/54 = 55.6%. Group B: 17/36 = 47.2%. Group A’s observed proportion is higher by about 8.33 percentage points using unrounded figures. Comparing 30 with 17 alone hides the unequal group sizes.
Case 11 — The misleading title
Question: A chart title says ‘X causes Group A behaviour’ on the basis of one two-way survey.
Worked answer: Replace it with a description such as ‘X selection by reported group in a 90-person survey’. A two-way frequency table describes the distribution of responses. It cannot, by itself, prove one variable causes the other. Numerical accuracy and causal language are different checks.
Case 12 — Unknown base from a rounded percent
Question: A report says 33% of an unspecified group chose water. Can the exact count be recovered?
Worked answer: No. Both the group size and the unrounded percentage are unknown. Many count-and-total pairs produce an approximately 33% figure. A correct answer identifies the missing denominator and the rounding ambiguity rather than inventing a participant count.
Case 13 — Exact count from an exact percent
Question: A group contains 80 people, and exactly 35% chose an option. Recover the count.
Worked answer: Calculate 0.35 × 80 = 28 people. Because the percentage is exact and the group total is known, the count is uniquely determined. A count must be a whole number; here the calculation produces one without rounding.
Case 14 — A non-integer result
Question: A displayed rate is 41% for a group of 82 people; multiplying gives 33.62. What should be reported?
Worked answer: Do not report 33.62 persons. The 41% may have been rounded, and the exact underlying frequency is not supplied. State the ambiguity, look for a rounding convention, and use an approximate count only if the question explicitly allows it.
Case 15 — A contradictory grand total
Question: The two row totals are 72 and 48, but a caption claims 125 complete records. Is the table consistent?
Worked answer: Not under a two-row exhaustive classification: 72 + 48 = 120. It could be that five submitted forms were invalid or excluded, but the caption must say so. Do not change 72 or 48 without evidence just to make 125 fit.
Case 16 — Multiple-response trouble
Question: Respondents could choose both water and juice, but a table assumes exactly one refreshment option.
Worked answer: The two choices are not necessarily mutually exclusive. A simple yes/no table for water may still be formed, but a water-versus-juice table would misclassify people who chose both. State how the survey categories must be defined before recovering missing cells.
Case 17 — A missing-response column
Question: One hundred twenty-five forms were received; five lacked the second survey answer. Can a two-way table with 120 complete records cover every form?
Worked answer: It can cover the 120 complete cases if that scope is made explicit. The five missing responses require separate disclosure or a missing-data category if analysing all forms. Never silently treat missing answers as ‘no’, because that changes category counts.
Case 18 — A probability within the full sample
Question: One of 120 complete records is selected uniformly at random. Forty belong to the target-met-and-water cell. Find the probability.
Worked answer: The probability is 40/120 = 1/3. The equally likely selection is from all 120, so the grand total is the denominator. Do not substitute the 72-person row total unless the random selection is restricted to that row.
Case 19 — A probability within one row
Question: One respondent is selected uniformly from the 72 who met the activity target. Forty of them chose water. Find the probability.
Worked answer: Use 40/72 = 5/9, approximately 0.556. The restriction ‘from the 72’ creates a new sample space. The same cell can support several valid fractions, depending on the described selection procedure.
Case 20 — A column comparison
Question: Of the 58 who chose water, 40 met the target and 18 did not. Which is the larger share within this column?
Worked answer: Forty of 58 is about 69.0%; eighteen of 58 is about 31.0%. Both shares use the same column denominator, so they add to 100% within rounding. Comparing them to percentages with row denominators would be an invalid mixture.
Case 21 — A reversed conditional
Question: A draft argues that P(A given X) must equal P(X given A), because both include the same 30 people.
Worked answer: The numerator can match while the denominators differ. In the 90-person example, P(A given X) = 30/47, whereas P(X given A) = 30/54. Conditional language is directional, even in a school arithmetic problem without formal probability notation.
Case 22 — A double-counted intersection
Question: Someone adds 72 who met the target and 58 who chose water and says 130 unique people were surveyed.
Worked answer: The intersection of 40 people appears in both counts. To find people who met the target or chose water, use 72 + 58 − 40 = 90. This is the count in the union. It does not imply the grand total exceeds 120.
Case 23 — Neither category
Question: From 120 records, 72 met the target, 58 chose water and 40 did both. How many did neither?
Worked answer: First calculate the union: 72 + 58 − 40 = 90. Therefore 120 − 90 = 30 chose neither category. This matches the bottom-right interior cell recovered by row and column subtraction, providing a useful second method.
Case 24 — A mistaken mean
Question: A learner averages 55.6% and 37.5% to estimate the overall water percentage across unequal groups.
Worked answer: A simple average is generally inappropriate because the group sizes are 72 and 48. Use the actual total 58/120 = 48.3% for the whole sample. Alternatively compute a weighted average of the unrounded row proportions using their group sizes.
Case 25 — A rounded display discrepancy
Question: Displayed group rates are 55.6% and 37.5%, and a calculator yields 18.1 percentage points. Why might the exact answer differ slightly?
Worked answer: The first displayed percentage is rounded from 40/72 × 100. Subtract the unrounded rates first: about 18.0556 percentage points, which rounds to 18.1 at one decimal place. Keep original fractions through intermediate work when a question is sensitive to precision.
Case 26 — The denominator moves unnoticed
Question: A report says ’40 out of 72, then 40 out of 58, therefore 80 out of 130 people’.
Worked answer: The two fractions refer to the same forty intersection participants counted from different perspectives. Adding them double counts those people and creates a meaningless total. Treat each conditional proportion separately and return to the 120 unique records when describing the full sample.
Case 27 — Possible bias
Question: The 120 volunteers signed up at a single school event. A headline says that their preferences represent all teenagers.
Worked answer: Voluntary event participants may not be representative of all teenagers. The table accurately records the participating group but cannot justify a population-wide statement without sampling evidence. An exam answer should distinguish the sample from the population the headline claims.
Case 28 — An alternative row definition
Question: One survey version defines ‘active’ as daily exercise, while another defines it as a weekly activity target. Can the tables be pooled directly?
Worked answer: Not without checking that the categories mean the same thing. Different definitions change who belongs in each row. A mathematically balanced pooled table may still answer an incoherent question. State the classification rule before merging records.
Case 29 — Rounding a frequency from 48.3%
Question: The entire sample has 120 people and a report says 48.3% chose water. Can you recover exactly 58?
Worked answer: A count of 58 gives 48.333…%, which rounds to 48.3%, but other nearby count-and-total combinations could also produce the displayed rate if the total were not fixed. With 120 fixed and ordinary rounding, 58 is the consistent integer. Show 58/120, and identify that the percentage was rounded.
Case 30 — A second table with 60 records
Question: Among sixty participants, forty use option A and twenty use option B. Of those using A, fifteen are in Group X; of those using B, eight are in Group X. Find group X total.
Worked answer: Group X has 15 + 8 = 23 people. The complementary row contains 40 − 15 = 25 and 20 − 8 = 12, total 37. The row totals 23 + 37 and column totals 40 + 20 both equal 60.
Case 31 — An incorrect axis for a bar chart
Question: A graph shows ‘number of water choosers’ for two groups of different sizes but titles it ‘water preference rates’.
Worked answer: Counts and rates are different quantities. To display rates, divide 40 by 72 and 18 by 48, then label the vertical axis with percentage. If a chart shows counts instead, name them honestly and provide group sizes so that readers do not infer a proportion incorrectly.
Case 32 — Deciding which result to state
Question: A 3-mark task asks you to complete the table and compare the percentages of water choosers in the two activity groups.
Worked answer: Use the supplied totals to recover 32, 18 and 30; verify margins; compute 40/72 = 55.6% and 18/48 = 37.5%; then state the first observed proportion is higher. Do not speculate on coronary outcomes, because none was measured.
Case 33 — Too many decimal places
Question: A candidate gives 55.5555555555% for a computed conditional proportion.
Worked answer: The repeating decimal reflects calculation, not the required level of answer precision. Where the question gives no special rounding instruction, a suitable three-significant-figure result is 55.6%. Preserve the exact fraction in working and give a readable final value.
Case 34 — Wrong group comparison
Question: The claim compares 40 of 72 with 18 of 120 and says the first group has much higher preference.
Worked answer: The first denominator is a subgroup, whereas the second is the whole sample. The groups are not comparable on those bases. Use 40/72 against 18/48 if comparing within-group preference rates, or choose another consistent comparison that matches the wording.
Case 35 — A missing cell makes a negative value
Question: A proposed table says a 40-person row contains 45 in its first cell and −5 in its second.
Worked answer: The algebraic result exposes inconsistent data: a cell count cannot be negative. Check the meaning of the row total, whether some values refer to different populations, and any transcription errors. Do not interpret negative people as a legitimate survey result.
A reproducible ten-minute table-check routine
For two minutes, read only the labels and circle the population described by each total. For the next three, fill the missing cells using row and column subtraction. Spend two minutes checking that row sums, column sums and the grand total agree. Use two minutes to write any percentages as verbal fractions with visible denominators. Reserve the last minute for a sentence that describes the result without inventing causation or a larger population.
If you make a mistake, do not just mark it wrong. Write the exact source of the mismatch: ‘I divided a subgroup count by the grand total’, ‘I added overlapping categories’, or ‘I treated a rounded percentage as an exact count’. These are different errors and require different repairs. Repeat one unfamiliar table the next day with new numbers and without looking at the worked answer.
Frequently asked questions
Does a two-way table prove one behaviour causes another? No. It organises records and permits comparisons; a causal conclusion needs different evidence. A fictional exercise about habits must never be converted into a judgement about a student’s health.
Are two-way tables themselves a separate SEC examination paper? No. This is a problem-solving application of K310 mathematics topics such as statistics, number work, percentages and interpretation. The official papers are Paper 1 and Paper 2, each 90 marks and 2 hours 15 minutes for 2027.
What is the fastest way to avoid the wrong denominator? Underline the words after ‘of’ or ‘among’, write that group’s count under the fraction line, and only then calculate the percentage.
Where does the series go next? English handles how to speak about a school’s survey results responsibly. The later EMS case combines Biology, Maths and English while leaving each specialised skill to its own guide.
Official syllabus and learning routes
Official: 2027 G3 SEC index and K310 Mathematics syllabus. Internal: Mathematics Hub, English Hub, PSLE route, Vol 0577 Biology. All survey examples here are invented for learning; they are not medical or public-health evidence.
