Quick answer: the denominator changes the story
When two groups have different sizes, counts alone can give a misleading picture. Before comparing the percentage of correct sign readings, identify exactly how many attempts belong in each condition. In a fictional habitat night-route audit, 35 of 50 attempts at a bright checkpoint succeed and 25 of 50 at a dim checkpoint succeed. The overall percentages are 70% and 50%. Yet the composition of participants differs so sharply that the opposite pattern appears when we compare experienced and new readers separately. This is a question about group mix, not an invitation to claim one lighting scheme is scientifically superior.
This Mathematics guide continues Volume 1333 on the eye and pupil reflex. That Biology article established why changes in pupil diameter do not prove a sign is readable. Here we learn to audit the arithmetic used in a proposed report. Our numbers are invented for examination-style practice and do not certify a real route, visual accessibility, or any individual.
The first rule is to write the fraction before the percentage. The second is to say which group the denominator represents. The third is to compare like with like and report any group differences before making a broad claim. These habits transfer to G3 Mathematics, Science graphs, English evidence questions and ordinary decisions beyond school.
At a glance: the central checks
• Identify the count, its complete reference group and its unit.
• Convert a valid fraction to a percentage: count divided by relevant total, then multiplied by 100.
• Distinguish a percentage point difference from a relative percentage change.
• Read two-way tables both across rows and down columns.
• Look for unequal subgroup sizes before interpreting an overall rate.
• Inspect chart baselines, labels, missing observations and sample design.
• A calculated rate is not by itself a causal explanation or an accessibility assessment.
The verified 2027 G3 SEC Mathematics route
The Singapore–Cambridge Secondary Education Certificate begins in 2027, replacing the former N- and O-Level certificates. SEAB identifies G3 Mathematics as syllabus K310, with 4052 as the older reference code. G3 refers to subject demand, not to Secondary 3 as a year level. The official SEAB list identifies the 2027 code, while the K310 syllabus document describes content and assessment.
K310 Paper 1 has about 26 short-answer questions, lasts two hours 15 minutes, carries 90 marks and is worth 50%. Paper 2 lasts two hours 15 minutes, carries 90 marks and is also 50%; it has nine to ten questions, including an extended real-world application at the end. Candidates answer all questions. The syllabus says that essential working must be shown, and non-exact numerical answers generally use three significant figures unless a question specifies otherwise; angles in degrees generally use one decimal place.
The Statistics and Probability strand includes analysing tables, graphs, misleading representations, statistical inference, measures of centre and spread, and elementary probabilities. Our focus is practical data interpretation and fair comparison. The invented night-route task is not an official past paper, nor a predicted SEC examination question.
Meet the four groups, not merely the two totals
The fictional route has two conditions, bright A and dim C. The records are divided into two categories: readers familiar with the signs (‘experienced’) and readers new to the task (‘new’). At bright A, 40 experienced attempts include 32 correct readings, while 10 new attempts include three correct readings. At dim C, 10 experienced attempts include nine correct readings, while 40 new attempts include 16 correct readings.
Write the table in words before calculating. Bright A: experienced 32 correct out of 40, new three correct out of 10, total 35 correct out of 50. Dim C: experienced nine correct out of 10, new 16 correct out of 40, total 25 correct out of 50. The number of attempts is 100 across both conditions; the number of correct readings is 60. Attempts are not necessarily unique people: the dataset does not say that they are.
The data are designed to expose a denominator trap. The groups under each condition are not made up in the same proportions. At bright A, 80% of attempts are from experienced readers. At dim C, only 20% are experienced. That compositional difference matters more than the first glance at the totals.
Start with four reliable fractions
For bright experienced attempts, the correct-reading rate is 32 divided by 40, or 80%. For bright new attempts, it is three divided by 10, or 30%. At dim C, the experienced rate is nine divided by 10, or 90%, and the new rate is 16 divided by 40, or 40%. Write all four fractions before interpreting the overall rate.
Within experienced attempts, the dim condition’s observed correct rate is 10 percentage points higher: 90% compared with 80%. Within new attempts, it is also 10 percentage points higher: 40% compared with 30%. Yet the bright condition’s overall rate is 70%, compared with 50% for dim. How can both statements be true? Because the bright group contains proportionally many more experienced attempts.
No arithmetic mistake is required for a misleading conclusion. A perfectly calculated overall rate can answer the wrong comparison question. That is why the student’s work must state both the denominator and the intended comparison.
Compare the same mix to expose the shift
To illustrate standardisation, imagine giving equal weight to experienced and new attempt rates in each condition. For bright A, an illustrative equal-weight average is (80% + 30%) ÷ 2 = 55%. For dim C, it is (90% + 40%) ÷ 2 = 65%. The observed direction now favours dim by 10 percentage points. These are hypothetical equal-mix rates, not the original observed total percentages.
Explain what changed and what did not. The four within-group success rates stayed fixed. Only their relative weights changed. An overall percentage is a weighted average of subgroup percentages. If the weights differ, the combined rate can differ for reasons unrelated to the condition the publicity statement highlights.
This is an advanced interpretation skill but uses accessible arithmetic. The student need not claim advanced statistical causation. State: the unbalanced mix could explain the apparent discrepancy, and a properly designed study is needed before a causal claim about lighting.
Why 20 percentage points is not a 20 percent change
The difference between the overall rates is 70% minus 50%, or 20 percentage points. A relative comparison uses a defined reference: measured against 50%, the difference of 20 points corresponds to a 40% relative increase. These are not interchangeable expressions. One uses the scale of percentages; the other divides the change by a baseline value.
Suppose an editor writes ‘bright is 20% better’. Which meaning do they intend? If they mean the overall observed proportion was 20 percentage points higher, they should say so. Even ‘better’ may be too strong, because the groups were not comparable. The most defensible statement is ‘the overall observed correct rate was 70% at A and 50% at C in this fictional dataset’.
Wording is a mathematical skill. Accurate computation with unclear language can leave the reader with an incorrect impression. G3 Mathematics demands interpretation of results in context, not just the ability to perform division.
Total attempts, participants and independent evidence
Do not assume that 100 attempts means 100 people. The case tells us 100 attempts, and nothing more. A person could have performed several trials. If a report switches to ‘100 visitors’, it has introduced a fact not supplied by the record. An accurate data table should label what was counted and how repeated observations were handled.
Likewise, ’60 correct attempts’ does not tell us the average reading speed, comfort or confidence. Error counts are only one part of a task. A statement about safety must be supported by appropriate evidence outside this arithmetic exercise.
The previous Science volume distinguished a measured pupil response from a conclusion about readability. This Mathematics volume makes an allied distinction: a percentage based on an attempt count is not automatically a percentage of unique people. Always ask, ‘What exactly is one unit in the denominator?’
Read two-way tables both directions
A two-way table can be read by condition or by experience category. If you calculate the share of experienced attempts within bright A, use 40/50 = 80%. If you calculate the share of bright attempts within all experienced attempts, use 40/50 = 80% coincidentally in this deliberately symmetric example; the same number disguises a different question.
To see why direction matters, calculate ‘correct among bright and new’ as 3/10 = 30%. Contrast that with ‘bright and new correct as a share of all bright attempts’: 3/50 = 6%. Both fractions use the same numerator and both are correct for their stated questions. They are not answers to the same question.
When a phrase contains ‘among’, ‘out of’, or ‘of all’, underline the specified reference group. Most denominator errors can be stopped before the calculator is used.
Make the misleading graph visible in words
Suppose a bar chart shows 70% and 50%, but its vertical axis begins at 45%. The visible bar heights appear very different. The numbers printed on the chart can be correct while the visual presentation exaggerates their contrast. The issue is not that every non-zero baseline is forbidden; it is whether the presentation could mislead the intended reader about magnitude.
Another chart omits the distinction between experienced and new attempts and labels the bars ‘readability by lighting’. That title suggests causation beyond the data. A better title is ‘Observed correct-reading rates in two differently composed attempt groups’. It is longer, but it does not hide the main limitation.
The PSLE Mathematics time-budget guide offers a reminder that method and checking matter under timed conditions. G3 adds more demanding data reasoning; a student should be able to explain why an apparently polished graph might tell the wrong story.
Work with counts before turning them into rates
The total correct attempts at bright A are 32 + 3 = 35. At dim C they are 9 + 16 = 25. Across the conditions there are 60 correct attempts. The incorrect attempts are 50 − 35 = 15 at A and 50 − 25 = 25 at C. That makes 40 incorrect attempts in total. The correct and incorrect counts sum to 100, which matches the total attempts and serves as an internal check.
This arithmetic is deliberately simple so the interpretive work becomes visible. In exams, wrong denominators can hide behind beautifully accurate sums. A reliable order is count, check, divide, label and interpret. Do not begin by copying a percentage from an eye-catching chart when the underlying group structure is supplied.
It is acceptable to show an exact fraction alongside the decimal. For example, 35/50 = 7/10 = 0.7 = 70%. Each representation refers to the same proportion only while the denominator remains the same group.
Introduce percentages without inventing causation
A strong mathematical conclusion would say: ‘Among 50 recorded attempts in each condition, the correct-reading proportions were 70% at bright A and 50% at dim C. However, bright A contained 40 experienced attempts compared with only 10 at dim C. Within each experience category the observed rate at dim C was higher, so the combined percentages do not isolate an effect of light level.’
The sentence does three jobs. It uses the numbers, exposes the composition and limits inference. It is therefore stronger than ‘the bright sign worked better’. The latter sounds easy to read but cannot be justified by the table.
Be especially careful with the word ‘because’. The data demonstrate differences in recorded rates. They do not establish why individuals made errors. A biological mechanism may be relevant, but the mathematics cannot by itself separate glare, practice, letter size and other factors.
What a better comparison might look like
A redesigned learning exercise could arrange comparable numbers of experienced and new attempts under both conditions, keep the displayed material constant and control the viewing procedure. The comparison must still consider repeated attempts from the same person and what an attempt actually measures. These are research-design issues, not calculations alone.
For a Mathematics paper, a plausible answer might identify the bias and explain how balancing categories reduces one source of distortion. It need not promise that balancing cures every problem. The value is in saying exactly which source of variation the proposed change targets.
An honest study can also report subgroup findings rather than collapsing all observations into one score. The goal is not to force an attractive single percentage. It is to make the underlying information usable for the decision being discussed.
Rates, decimal precision and final answers
The K310 syllabus asks students to use stated accuracy instructions, and generally gives non-exact numerical answers to three significant figures unless the question sets another standard. Exact percentages such as 70%, 50%, 80% and 90% need not be embellished with artificial decimals. If a calculation gives 84.375%, decide whether the problem asks for an exact fraction, a percentage to one decimal place, or another appropriate representation.
A fraction can sometimes be clearer than an unwarranted decimal. Reporting ‘three of ten’ immediately reveals the small subgroup. The statement ‘30.000%’ may make the same result sound more precise without adding any evidence.
Carry working values accurately and round at the end. If you round several intermediate percentages and then recombine them, discrepancies can appear. The reader should always be able to trace a quoted rate back to its numerator and denominator.
The last question is usually an interpretation question
In a real-world application, Mathematics is not complete when the calculator shows an answer. Ask whether the percentage uses the right group, whether the comparison is fair, and whether the context has changed the meaning of a unit. If the answer is a proportion, does the number fall between zero and one before conversion to a percentage? If the answer is a count, must it be a whole number?
The useful habit from Primary school is to check reasonableness. At G3, add a model check: did the mathematical representation preserve the relevant distinctions? A two-way table preserves experience categories that an overall average can conceal.
For further support, the Secondary 4 Mathematics error-ledger guide gives a method for finding the first incorrect step rather than calling every loss ‘careless’. This is the approach to use when reviewing the questions below.
A timed plan for this data question
Spend a short initial period annotating group definitions and creating the four fractions. Complete straightforward totals next. Then answer the interpretive prompts with both the calculated results and a clear limitation. Reserve a last check for units, denominators and phrasing. This is a method, not an official minute allocation.
Do not race through the interpretation because the arithmetic looks elementary. More advanced questions often hide the difficulty in an apparently simple chart or in the phrase ‘as a percentage of’. The first ten seconds spent identifying the reference group can prevent a long chain of wrong calculations.
Students can practise by covering the worked responses and completing an entire sequence with no hints. After marking, classify every error as a counting error, a reference-group error, a representation error or an inference error. Each category needs a different repair.
An honest conclusion fits in one paragraph
The fictional records show 35/50 correct attempts (70%) under bright A and 25/50 (50%) under dim C. They also show that experienced attempts form 40/50 of bright records but only 10/50 of dim records. Within both experienced and new groups, the observed dim-condition percentages are higher. The combined comparison is therefore sensitive to the different group mix. These records alone do not establish that either checkpoint provides better visibility for the public.
That paragraph is more informative than a ranking. It shows how many observations exist, what they measure, why the aggregate is vulnerable and what cannot be concluded. Students should aim to write such short, accurate explanations after doing the longer working.
The next English volume will ask how to turn this corrected statistical understanding into a trustworthy public-facing report. An accessible written statement must not erase the caution discovered in the Mathematics exercise.
Worked SEC-style data checks: practise before reading the answers
These questions are original, invented for this guide and aligned to relevant K310 skills. They are not SEAB past-paper reproductions. Record the denominator explicitly for each percentage. After reading a model answer, repeat the question with new figures to test transfer rather than memorising the original number.
Worked check 1: Find the bright total
Problem. Add the 32 correct experienced attempts and the three correct new attempts.
Worked reasoning. There are 35 correct bright attempts out of 50 attempts. This count is the numerator for the overall bright success rate.
Error to avoid. Do not use 32 alone unless the question names the experienced subgroup.
Transfer task. Calculate the corresponding dim total.
Worked check 2: Find the dim total
Problem. Add nine experienced correct attempts and 16 new correct attempts.
Worked reasoning. The dim correct total is 25. The denominator is 10 + 40 = 50 attempts.
Error to avoid. The fact that 16 exceeds nine does not mean new readers have a higher within-group rate.
Transfer task. Calculate both dim subgroup percentages.
Worked check 3: Bright experienced rate
Problem. Compute 32 correct attempts out of 40 experienced attempts.
Worked reasoning. 32/40 = 0.8 = 80%. All 40 attempts in this condition-category cell form the reference group.
Error to avoid. Using 50 would answer ‘share of all bright attempts’ instead.
Transfer task. Explain why the two denominators correspond to different questions.
Worked check 4: Bright new rate
Problem. Compute three correct attempts out of ten new attempts.
Worked reasoning. 3/10 = 30%. The result is a proportion among new attempts at bright A.
Error to avoid. It is not 3% simply because the numerator equals three.
Transfer task. State the incorrect-attempt rate for this subgroup.
Worked check 5: Dim experienced rate
Problem. Compute nine correct attempts out of ten experienced attempts.
Worked reasoning. 9/10 = 90%, which is 10 percentage points above the corresponding bright experienced rate.
Error to avoid. Do not claim a lighting effect without considering the sampling arrangement.
Transfer task. Give the exact fraction of unsuccessful dim experienced attempts.
Worked check 6: Dim new rate
Problem. Compute 16 correct attempts out of 40 new attempts.
Worked reasoning. 16/40 = 40%. The denominator must be the 40 new attempts made at dim C.
Error to avoid. Comparing 16 to three as raw counts ignores subgroup size.
Transfer task. Compare the correct-reading rates for new attempts under both conditions.
Worked check 7: Bright overall rate
Problem. Compute 35 correct attempts out of 50 bright attempts.
Worked reasoning. 35/50 = 70%. This describes recorded attempts under the bright condition.
Error to avoid. It does not say that 70% of all habitat residents could read the sign.
Transfer task. Name the exact population observed.
Worked check 8: Dim overall rate
Problem. Compute 25 correct attempts out of 50 dim attempts.
Worked reasoning. 25/50 = 50%. The denominator includes both experience categories at dim C.
Error to avoid. The percentage should not be calculated as 25/100 unless the reference is all attempts.
Transfer task. State the overall incorrect rate for dim C.
Worked check 9: Total correct attempts
Problem. Combine bright and dim correct counts.
Worked reasoning. 35 + 25 = 60 correct attempts across 100 recorded attempts. Overall this is 60%.
Error to avoid. A pooled percentage is not a description of a typical participant.
Transfer task. Explain why unique people cannot be counted from this record.
Worked check 10: Total incorrect attempts
Problem. Deduce wrong readings without extra data.
Worked reasoning. Bright contributes 15 and dim contributes 25 incorrect attempts, a total of 40.
Error to avoid. Check that correct plus incorrect equals all recorded attempts.
Transfer task. Give the bright incorrect percentage.
Worked check 11: Experienced attempts total
Problem. Count experienced attempts across both conditions.
Worked reasoning. 40 + 10 = 50 experienced attempts. There are 32 + 9 = 41 correct experienced attempts overall.
Error to avoid. This pooled experienced rate mixes both lighting conditions.
Transfer task. Calculate 41/50 and give its limited meaning.
Worked check 12: New attempts total
Problem. Count new attempts across both conditions.
Worked reasoning. 10 + 40 = 50 new attempts, with three + 16 = 19 correct; pooled rate 19/50 = 38%.
Error to avoid. The same group total does not imply the same distribution by condition.
Transfer task. State where most new attempts occurred.
Worked check 13: Group composition at A
Problem. What fraction of bright attempts comes from experienced readers?
Worked reasoning. 40/50 = 80%, so bright A is dominated by experienced attempts.
Error to avoid. This is a composition rate, not a correct-reading rate.
Transfer task. State the share of new attempts at bright A.
Worked check 14: Group composition at C
Problem. What fraction of dim attempts comes from experienced readers?
Worked reasoning. 10/50 = 20%, so dim C has a very different group mix from A.
Error to avoid. Comparing two overall rates without noting this 60-point compositional difference is risky.
Transfer task. Find the share of new attempts at C.
Worked check 15: Percentage points
Problem. Compare 70% and 50% correctly.
Worked reasoning. The observed difference is 20 percentage points. That is a subtraction of rates expressed in percent.
Error to avoid. A 20-percentage-point gap is not the same statement as a 20% relative increase.
Transfer task. Calculate the relative difference using 50% as baseline.
Worked check 16: Relative increase
Problem. Use 50% as the reference for a change to 70%.
Worked reasoning. (70 − 50)/50 × 100% = 40% relative increase, in this purely arithmetic comparison.
Error to avoid. ‘40% better’ would still overclaim about lighting because groups are different.
Transfer task. Show the corresponding decrease from 70% down to 50%.
Worked check 17: Relative decrease
Problem. Use 70% as the reference for a change down to 50%.
Worked reasoning. (70 − 50)/70 × 100% = approximately 28.6% relative decrease.
Error to avoid. Relative change depends on which reference is chosen.
Transfer task. Explain why the percentage-point difference remains 20.
Worked check 18: Equal-mix bright rate
Problem. Average the two bright subgroup rates with equal weights.
Worked reasoning. (80% + 30%)/2 = 55%. This is a hypothetical 50:50 composition, not the recorded overall 70%.
Error to avoid. Do not relabel an adjusted example as observed data.
Transfer task. Compute the corresponding hypothetical dim rate.
Worked check 19: Equal-mix dim rate
Problem. Average dim subgroup rates with equal weights.
Worked reasoning. (90% + 40%)/2 = 65%, ten percentage points above the equal-mix bright value.
Error to avoid. The adjusted contrast does not prove causality; it only removes this one compositional imbalance.
Transfer task. Explain what stayed constant during the adjustment.
Worked check 20: Why the ranking reverses
Problem. Explain why bright looks higher overall but lower within both categories.
Worked reasoning. Bright contains many more experienced attempts. Experienced readers have higher success proportions in both conditions, shifting the combined result upward.
Error to avoid. Do not accuse the arithmetic of error: the totals were calculated correctly.
Transfer task. State the limitation in one calm sentence.
Worked check 21: Conditional denominator
Problem. What is ‘correct among new attempts at dim C’?
Worked reasoning. 16/40 = 40%. ‘Among new’ specifies the restricted denominator.
Error to avoid. Using 16/50 would be share of all dim attempts, a different quantity.
Transfer task. Compute correct new attempts as a share of all dim attempts.
Worked check 22: Share of all dim
Problem. What is ‘correct new attempts as a share of all dim attempts’?
Worked reasoning. 16/50 = 32%. This combines the new-category and success conditions in the numerator.
Error to avoid. It must not be described as the new subgroup’s success rate.
Transfer task. Write the phrase ‘out of all dim attempts’ before a 32% result.
Worked check 23: Share of all 100
Problem. What fraction of all recorded attempts are correct bright experienced attempts?
Worked reasoning. 32/100 = 32%. The reference is every attempt in the complete dataset.
Error to avoid. The subgroup success rate was 80%; the numbers answer different questions.
Transfer task. Compare both numbers in one sentence.
Worked check 24: Correct bright share of all correct
Problem. Among all 60 correct attempts, how many occurred at bright A?
Worked reasoning. 35/60 = 7/12, approximately 58.3%. The denominator is only correct attempts.
Error to avoid. It is neither the bright success rate nor the bright share of all attempts.
Transfer task. Calculate 25/60 for dim and check they sum to 100%.
Worked check 25: Part-whole check
Problem. Why must each cell count be no larger than its row and column total?
Worked reasoning. A subgroup is part of a larger collection; its count cannot exceed that containing total in a valid nonnegative frequency table.
Error to avoid. A cell greater than its row total signals transcription or definition trouble.
Transfer task. Test the 32-count against totals of 40 and 50.
Worked check 26: Percent outside range
Problem. A calculator gives 140% correct attempts. What should you check?
Worked reasoning. For correct outcomes divided by all attempts in the same group, a valid rate must be between 0% and 100%. Revisit numerator and denominator.
Error to avoid. Some percentage changes can exceed 100%, but an ordinary subset proportion cannot.
Transfer task. Explain why the context determines the bound.
Worked check 27: Units: attempts not people
Problem. Could you write ’60 people succeeded’ from the 60 correct attempts?
Worked reasoning. No. Repeated attempts may come from the same person. Report the unit actually counted.
Error to avoid. Switching unit words silently changes the claim even when the numeral is unchanged.
Transfer task. Draft a truthful public sentence using ‘attempts’.
Worked check 28: Unreported attempts
Problem. Five additional dim attempts were not scored. What happens to the reported 50%?
Worked reasoning. The rate 25/50 describes scored attempts. Without outcomes for five further attempts, a total attempted-score percentage is not fully known.
Error to avoid. Do not silently mark missing attempts incorrect or successful.
Transfer task. Give the lowest and highest possible success rates out of 55.
Worked check 29: Missing-data bounds
Problem. Find the extremes for 25 successes plus five unscored dim attempts.
Worked reasoning. If none of five succeeds the rate is 25/55 ≈45.5%. If all five succeed it is 30/55 ≈54.5%.
Error to avoid. The original 50% is a rate among scored observations, not a proven all-attempt rate.
Transfer task. Explain which additional record would narrow this range.
Worked check 30: Display baseline
Problem. A bar chart uses a vertical axis beginning at 45%. Why might it mislead?
Worked reasoning. The difference between 50% and 70% occupies a large portion of the cropped scale and can appear disproportionately dramatic.
Error to avoid. A bar chart’s appearance is not solely determined by valid labels.
Transfer task. Propose a clearer visual baseline or annotate the scale prominently.
Worked check 31: Misleading chart title
Problem. A chart is titled ‘Brightness causes better reading’. Repair it.
Worked reasoning. ‘Observed correct-reading rates by checkpoint, with different group mixes’ describes the data without inventing causation.
Error to avoid. The title should not outrun what the table can show.
Transfer task. Propose a short subtitle explaining experience-category imbalance.
Worked check 32: A two-way table heading
Problem. A student writes ‘success’ without a count definition. Improve it.
Worked reasoning. Use ‘number of correct sign-reading attempts’ and label the relevant row and column categories.
Error to avoid. Ambiguous headings make it impossible to verify a percentage’s denominator.
Transfer task. Label a second column ‘total attempts’.
Worked check 33: Question asks column percentage
Problem. The prompt says ‘Of all new attempts, what share occurred at C?’
Worked reasoning. 40/50 = 80%. The denominator is all 50 new attempts across conditions.
Error to avoid. New-at-C success is 40%, a different question involving correct outcomes.
Transfer task. Underline the reference phrase before dividing.
Worked check 34: Question asks row percentage
Problem. The prompt says ‘Of all C attempts, what share were new?’
Worked reasoning. 40/50 = 80%. It happens to match the previous answer here, but the reference set differs.
Error to avoid. Equal numerals do not prove two questions mean the same thing.
Transfer task. Explain the two denominators in words even when both equal 50.
Worked check 35: A ratio of group sizes
Problem. Express experienced:new counts at bright A as a ratio.
Worked reasoning. 40:10 simplifies to 4:1. At dim C it is 10:40 or 1:4.
Error to avoid. Ratios compare counts, not success proportions.
Transfer task. Explain why these ratios immediately flag imbalance.
Worked check 36: An alternative allocation
Problem. What if both conditions had 20 experienced and 20 new attempts?
Worked reasoning. That proposed design would balance these two category counts, making an overall rate less vulnerable to different mixes of experience.
Error to avoid. Balancing group sizes alone would not guarantee matching illumination or elimination of repeated-person bias.
Transfer task. Name one other factor requiring control.
Worked check 37: Rounding at the end
Problem. Why should you retain 35/60 before reporting its percentage?
Worked reasoning. The exact fraction avoids premature rounding; convert once at the reporting stage to approximately 58.3% when one decimal place is suitable.
Error to avoid. Rounding intermediate values repeatedly can introduce avoidable error.
Transfer task. Express 25/60 to the same accuracy.
Worked check 38: Check the pooled figure
Problem. A student averages 70% and 50% to get 60%. Is that valid here?
Worked reasoning. Yes, because both conditions have 50 recorded attempts, so a simple average of the two percentages equals the pooled 60%.
Error to avoid. Averaging percentages is not generally valid if group denominators differ.
Transfer task. Try 70% of 20 and 50% of 80 and recompute the pooled rate.
Worked check 39: Unequal-denominator counterexample
Problem. For 20 attempts at 70% and 80 attempts at 50%, find the total rate.
Worked reasoning. There are 14 + 40 = 54 successes out of 100, or 54%, not the simple average of 60%.
Error to avoid. Averaging percentages without weights discards sample-size information.
Transfer task. Explain which denominator acts as the weight.
Worked check 40: Comparison with Biology
Problem. What does a change from pupil diameter 3.0 mm to 5.0 mm establish?
Worked reasoning. The diameter increased by 2.0 mm in the fictional observation. It is not a success rate for reading.
Error to avoid. A biological response and a reading outcome are separate variables.
Transfer task. Name a practical check that does not make a medical claim.
Worked check 41: Mathematical model boundary
Problem. Can the 70% and 50% totals tell us which lamp should be installed?
Worked reasoning. Not by themselves. They are rates from differently composed records, and installation choices require proper task, design and safety evidence.
Error to avoid. Mathematics can reveal imbalance; it cannot turn flawed sampling into certainty.
Transfer task. State the narrowest recommendation supported by these figures.
Worked check 42: Error diagnosis
Problem. A learner calculates 16/10 for the new dim group. Identify the first failure.
Worked reasoning. The denominator 10 belongs to experienced dim attempts, not new dim attempts. The correct new dim fraction is 16/40.
Error to avoid. Correct the reference group before checking arithmetic.
Transfer task. Describe how a table highlight prevents this swap.
Worked check 43: Error diagnosis, second case
Problem. A student uses the 100-attempt total for every subgroup rate. What is wrong?
Worked reasoning. Each subgroup rate must divide by the total attempts within that subgroup. The 100 total is only appropriate for a share of the complete dataset.
Error to avoid. The correct numerator cannot rescue the wrong denominator.
Transfer task. Correct the bright new success rate and name its reference set.
Worked check 44: Context word choice
Problem. Is the sentence ‘40% of new readers succeeded at C’ precise?
Worked reasoning. Not quite: the records count attempts, not necessarily unique readers. ‘40% of recorded new-category attempts at C were correct’ preserves the data’s unit.
Error to avoid. Small wording changes can remove a large inferential error.
Transfer task. Write an equally precise sentence for experienced attempts at C.
Worked check 45: One-sentence summary
Problem. Give a concise conclusion that includes the key caution.
Worked reasoning. The bright group’s overall correct-attempt rate was higher, but its experience mix differed strongly, and both subgroups showed higher observed rates under dim C.
Error to avoid. This is an interpretation of fictional data, not a universal visibility finding.
Transfer task. Explain the difference to an English writing student.
Worked check 46: Advance to explanation
Problem. Why is a group’s apparent ranking an unsafe headline here?
Worked reasoning. A headline compresses the table into one number and hides a major compositional difference. The arithmetic needs a stated reference group and caveat.
Error to avoid. A headline should not turn association into proof of cause.
Transfer task. Draft a 16-word accurate headline.
Worked check 47: Exit ticket
Problem. How do you check any rate claim in three steps?
Worked reasoning. Find the numerator, name the denominator and verify the resulting statement refers to the same unit and group. Then compare like categories.
Error to avoid. A well-computed figure can still be attached to the wrong description.
Transfer task. Explain the method to a Primary 6 student.
Review and next route
When you finish, put away your notes and reconstruct the two-by-two comparison from its counts. Explain why the 70%-versus-50% overall ranking coexists with 80%-versus-90% and 30%-versus-40% subgroup comparisons. You should be able to use three sentences rather than twenty: one for the numbers, one for the different group mix and one for the inference limit.
Next, turn the analysis into a clear report for readers who will not inspect the whole working. The English Paper 1 companion will ask which figures must be included, which terms such as ‘visitors’ cannot be justified and how to write a balanced recommendation without implying a scientific safety test. This follows the evidence habit developed in Biology Volume 1333.
For revision, return to the EMS learner’s guide and the PSLE Mathematics foundation route. The examination scheme and content checked here are from the official K310 syllabus for 2027, as listed by SEAB on 14 November 2026. For later examination years, check the latest official syllabus rather than assuming the code or scheme will be unchanged.
