Level 3 Statistical reports practice paper 2026
Question One
Report 1: Phones in bed
This report is fictional. It was written for this practice paper, and its organisations and figures are made up.
Kōtuku Research surveyed a random sample of 1100 New Zealand adults by phone and online in August, asking about their phone habits.
Main findings: 57% of New Zealand adults use their phone in bed every night.
Asked what they do first thing in the morning, 44% said they check their phone, 31% said they make a hot drink, and the rest gave other answers.
Of the 400 adults aged under 35 in the sample, 66% use their phone in bed every night, compared with 52% of the 700 adults aged 35 and over.
Each person answered the question: "Like most New Zealanders, do you use your phone in bed every night?"
(i) Use the rule of thumb to calculate the margin of error for the whole survey of 1100 adults.
(ii) Explain why a margin of error is needed when reporting the results of this survey.
Hint
For one proportion from a random sample, the rule of thumb margin of error is 1√n.
Answer
(i) MOE = 1√1100 = 0.0302, about 3.0%. (ii) A different random sample of 1100 adults would give a slightly different percentage, so the margin of error shows how far the sample result could be from the result for all New Zealand adults.
Step by step
- (i) The sample size is n = 1100.
- Rule of thumb: MOE = 1√n = 1√1100 = 0.0302.
- As a percentage, the margin of error is about 3.0%.
- (ii) The 57% comes from one sample of 1100 adults, not from every New Zealand adult.
- Another random sample of 1100 adults would give a slightly different percentage, because of natural sample to sample variation.
- The margin of error tells readers how far the survey percentage could reasonably be from the true percentage of all New Zealand adults who use their phone in bed every night.
Full working
- n = 1100
- MOE = 1√1100 = 0.0302 = 3.0%rule of thumb for one proportion
- MOE needed: each random sample of 1100 gives a slightly different percentage
Construct an approximate 95% confidence interval for the proportion of New Zealand adults who use their phone in bed every night, and write an inference statement in context.
Hint
The confidence interval is the survey percentage plus and minus the margin of error from (a).
Answer
57% ± 3.0% gives 54.0% to 60.0%. We are pretty sure that between 54.0% and 60.0% of all New Zealand adults use their phone in bed every night.
Step by step
- Sample percentage: 57%. Margin of error from (a): 3.0%.
- Lower limit: 57% − 3.0% = 54.0%. Upper limit: 57% + 3.0% = 60.0%.
- Inference: we are pretty sure that the proportion of all New Zealand adults who use their phone in bed every night is somewhere between 54.0% and 60.0%.
- A good statement names the population (New Zealand adults) and the variable (using a phone in bed every night).
Full working
- 57% ± 3.0%
- CI = [54.0%, 60.0%]
- Pretty sure 54.0% to 60.0% of all NZ adults use their phone in bed every night
The report claims: "More New Zealand adults check their phone first thing in the morning than make a hot drink."
Use an appropriate confidence interval to evaluate this claim.
Hint
Both percentages come from the same question answered by the same 1100 people, so this is a comparison within one group. Its rule of thumb margin of error is 2√n.
Answer
Difference 44% − 31% = 13%. MOE = 2√1100 = 6.0%. CI 13% ± 6.0% = 7.0% to 19.0%. The interval is entirely positive, so the claim is supported: we are pretty sure that the proportion of NZ adults who check their phone first is between 7.0% and 19.0% higher than the proportion who make a hot drink first.
Step by step
- The two percentages are answers to the same question from the same sample, so this is a difference within one group.
- Difference: 44% − 31% = 13%.
- Rule of thumb for a difference within one group: MOE = 2√n = 2√1100 = 0.0603, about 6.0%.
- Confidence interval: 13% ± 6.0% = 7.0% to 19.0%.
- Inference: we are pretty sure that the proportion of all New Zealand adults who check their phone first thing in the morning is between 7.0% and 19.0% higher than the proportion who make a hot drink first.
- The whole interval is above 0, so the survey supports the claim.
- Classic mistake: using 1√n or the between groups rule here. The same people answered once, so the two percentages are linked.
Full working
- difference = 44% − 31% = 13%
- MOE = 2√1100 = 0.0603 = 6.0%within one group
- CI = 13% ± 6.0% = [7.0%, 19.0%]
- all positive, so the claim is supported
A news website used the report to claim: "Younger New Zealanders are more likely to use their phone in bed every night than older New Zealanders."
Use an appropriate confidence interval to evaluate this claim, and write an inference statement in context.
Hint
These are two separate groups of different sizes. Find 1√n for each group, average them, then multiply by 1.5.
Answer
Difference 66% − 52% = 14%. MOE for under 35s = 1√400 = 5.0%, for 35 and over = 1√700 = 3.8%; average 4.4%; 1.5 × average = 6.6%. CI 14% ± 6.6% = 7.4% to 20.6%. We are pretty sure that the proportion of NZ adults under 35 who use their phone in bed every night is between 7.4% and 20.6% higher than for adults aged 35 and over. The interval is entirely positive, so the claim is supported.
Step by step
- The under 35s and the 35 and overs are two separate groups, so this is a difference between two groups.
- Difference: 66% − 52% = 14%.
- Under 35s: 1√400 = 0.0500. Aged 35 and over: 1√700 = 0.0378.
- Average MOE = 0.0500 + 0.03782 = 0.0439.
- Rule of thumb for a difference between two groups: MOE = 1.5 × 0.0439 = 0.0658, about 6.6%.
- Confidence interval: 14% ± 6.6% = 7.4% to 20.6%.
- Inference: we are pretty sure that the proportion of New Zealand adults under 35 who use their phone in bed every night is between 7.4% and 20.6% higher than the proportion of adults aged 35 and over.
- The whole interval is above 0, so the survey supports the website's claim.
Full working
- difference = 66% − 52% = 14%
- 1√400 = 0.0500 and 1√700 = 0.0378
- average = 0.0439; MOE = 1.5 × 0.0439 = 0.0658 = 6.6%between two groups
- CI = 14% ± 6.6% = [7.4%, 20.6%]
- all positive, so the claim is supported
Look at the wording of the question in Report 1, and at how the survey was carried out.
Discuss TWO possible sources of non-sampling error in this survey, and how each could have affected the 57% result.
Hint
Think about what the words "Like most New Zealanders" suggest, and whether people tell the truth about habits they might feel judged for.
Answer
For example: (1) The question is leading: "Like most New Zealanders" suggests that the usual answer is yes, so some people may say yes to fit in, making 57% too high. (2) The answers are self-reported: people may not remember accurately or may not want to admit a habit they see as unhealthy, so the true figure could be higher or lower than 57%. Other valid sources: non-response (people who use their phone a lot may be more willing to answer a phone or online survey), or the survey asking about "every night" when habits vary.
Step by step
- Non-sampling errors are errors that come from how the survey was run or worded, not from which people happened to be chosen. A bigger sample does not fix them.
- Question wording: "Like most New Zealanders, do you..." tells people what the majority does. This leading wording pushes people towards yes, so the 57% is likely to be an overestimate.
- Self-reporting: people report their own habits. Some may feel embarrassed and say no, which would make 57% an underestimate; others may not remember accurately.
- Non-response: the survey was by phone and online. People who did not answer may differ from those who did, for example heavy phone users might be more likely to respond online.
- For Excellence, each source must be linked to this survey AND to how it could change the 57%.
Full working
- leading question "Like most New Zealanders": pushes answers towards yes, 57% likely too high
- self-reported habit: embarrassment or poor recall, 57% could be too low or too high
- accept non-response or other valid sources linked to this survey
Question Two
Report 2: A walk after lunch
This report is fictional. It was written for this practice paper, and its organisations and figures are made up.
Researchers at a university in Hamilton wanted to know whether a short walk after lunch improves concentration in the afternoon. They put up posters asking for student volunteers, and 120 students took part.
Each student was randomly allocated to one of two groups of 60. After lunch, one group went for a 10-minute walk outside; the other group sat quietly in a lounge for 10 minutes.
All students then did the same 15-minute concentration test, at 1:30 pm on the same day. The tests were marked by a research assistant who did not know which group each student was in.
The mean test score was 74 for the walking group and 68 for the sitting group. The researchers concluded: "A short walk after lunch improves afternoon concentration."
The report also included a second finding from a questionnaire: students who said they walk to classes most days had a higher mean grade than students who said they drive or take the bus.
(i) Explain why the walking study is an experiment, not an observational study.
(ii) Identify the explanatory variable and the response variable in this experiment.
Hint
Who decided which students walked? In an observational study, the researchers only record what people already do.
Answer
(i) The researchers decided, by random allocation, which students walked and which sat, so they deliberately changed the explanatory variable. (ii) Explanatory variable: what the student did after lunch (a 10-minute walk or sitting quietly). Response variable: the student's score on the concentration test.
Step by step
- (i) In an experiment, the researchers impose a treatment. Here they decided which students went for a walk and which sat quietly.
- In an observational study they would only have recorded what students chose to do after lunch.
- (ii) The explanatory variable is the one the researchers changed: walking for 10 minutes or sitting quietly for 10 minutes.
- The response variable is what they measured: each student's score on the concentration test.
- Generic statements such as "it has a treatment" are not enough; link them to this study.
Full working
- researchers allocated walk or sit, so it is an experiment
- explanatory: walk 10 minutes or sit 10 minutes after lunch
- response: concentration test score
Explain why the researchers randomly allocated the students to the two groups.
Hint
What other things, apart from walking, could affect a student's concentration score? What does random allocation do to them?
Answer
Random allocation balances out other variables that could affect concentration, such as how much sleep students had, what they ate, their fitness or their usual ability, so the two groups are alike apart from the walk. Then a difference in mean scores can be put down to the walk rather than to differences between the groups.
Step by step
- Many things could affect a concentration score: sleep the night before, what a student ate, caffeine, fitness, stress, or general ability.
- If students chose their own group, the keen or fitter students might choose to walk, and they might also concentrate better anyway.
- Random allocation spreads these other variables roughly evenly across the two groups.
- So the groups differ (on average) only in the treatment, and a difference in mean scores can be attributed to walking after lunch.
Full working
- other variables: sleep, food, caffeine, fitness, ability
- random allocation balances them across the two groups
- so a difference in mean score can be attributed to the walk
The tests were marked by a research assistant who did not know which group each student was in, and every student did the same test at the same time.
Explain how each of these features helps the experiment.
Hint
Could a marker who expects the walkers to do better mark them differently? What could change if the groups did the test at different times of day?
Answer
Blind marking stops the marker's expectations (for example, hoping walking helps) from affecting the scores they give. Using the same test at the same time keeps those conditions the same for both groups, so a difference in mean scores is not caused by an easier test or by tiredness at a different time of day.
Step by step
- Blinding the marker: if the marker knew who walked, they might, even without meaning to, give walkers the benefit of the doubt. Not knowing the groups removes this bias.
- Same test: if the groups did different tests, one test might be easier, and that would show up as a difference in scores.
- Same time: concentration changes during the afternoon. Testing everyone at 1:30 pm on the same day keeps this the same for both groups.
- Both features control other variables, so a difference in the means is more likely to be caused by the walk.
Full working
- blind marker: expectations cannot bias the scores
- same test, same time: test difficulty and time of day are the same for both groups
- so the difference in means is more likely due to the walk
A health magazine reported the experiment with the headline: "A walk after lunch will boost concentration for all New Zealanders."
Discuss ONE issue with extending the results of this experiment to support the magazine's claim.
Hint
Who were the 120 participants, and how were they found? Who does "all New Zealanders" include?
Answer
For example: the participants were university students who volunteered after seeing a poster, not a random sample of New Zealanders. Volunteers for a walking study may be fitter or more keen on exercise, and university students are mostly young adults. Older people, children or people with health conditions may respond to a walk differently, so the results can only be extended to students like these volunteers, not to all New Zealanders.
Step by step
- Random allocation lets us make a causal claim, but only for the people in the experiment and people like them.
- The 120 students were volunteers who answered a poster at one university in Hamilton. They were not randomly selected from New Zealanders.
- They are likely to be young adults, and volunteers for a walking study may be more interested in exercise or fitter than average.
- Children, older people, shift workers or people with health conditions may react to a walk differently.
- So the experiment does not support extending the result to all New Zealanders. Other valid issues: only one afternoon and one test were used, and a 10-minute walk outside may depend on the weather.
- Sample size is not the issue here; do not argue that 120 is too small.
Full working
- participants: volunteer university students in Hamilton, not a random sample of NZers
- young, possibly fitter or keener on exercise than average
- results extend only to students like these, not all New Zealanders
The report's second finding was that students who walk to classes most days had a higher mean grade than students who drive or take the bus. A student newspaper concluded: "Walking to class makes you get better grades."
Evaluate this conclusion.
Hint
Did the researchers decide who walked to class? What else is different about students who walk to class?
Answer
The conclusion is not justified. This finding comes from a questionnaire, so it is an observational study: students chose how they travel. Students who walk probably live closer to campus (for example in halls), may have more time to study and may differ in age, work hours or income from students who drive. Any of these could explain the higher grades. An observational study can show an association but not that walking causes better grades.
Step by step
- In the questionnaire, nobody was allocated to walk or drive. Students chose their own way of travelling, so this is an observational study.
- Observational studies can show that two variables are related, but they cannot show cause and effect.
- Possible other variables: students who walk often live in halls near campus, so they spend less time travelling and more time studying; students who drive may be older and working part time.
- These other variables, not the walking, could explain the higher mean grade.
- Also, the travel method was self-reported, so some answers may be inaccurate.
- Conclusion: the data show an association, but the newspaper's causal claim is not supported. An experiment with random allocation would be needed to test it.
Full working
- questionnaire: students chose how they travel, so observational
- other variables: living near campus, study time, age, work hours
- association only; the causal claim is not supported
Question Three
Report 3: The Te Awa cycleway
This report is fictional. It was written for this practice paper, and its organisations and figures are made up.
Te Awa District Council is deciding whether to build a cycleway beside its main river. The council wanted to know what the district's adult residents think.
Survey A: the council put an online form on its Facebook page for four weeks. 2500 people filled it in, and 64% supported the cycleway. A council press release said: "With 2500 responses, our margin of error is only 2%."
Survey B: a research company phoned a random sample of 600 adults on the electoral roll for the district. 47% supported the cycleway, and 12% said they cycle to work at least once a week.
For Survey A, identify the target population and the sampling frame, and describe ONE way they do not match.
Hint
The target population is who the council wants to know about. The sampling frame is the list or source the responses could come from.
Answer
Target population: adult residents of Te Awa District. Sampling frame: people who see the council's Facebook page. They do not match: many residents do not use Facebook or do not follow the council, so they could not take part, while people who live outside the district, or are under 18, could still fill in the form.
Step by step
- Target population: all adult residents of Te Awa District.
- Sampling frame: people who see the council's Facebook page during the four weeks.
- Mismatch 1: residents who do not use Facebook, or do not follow the council, have no chance to respond. Older residents may be under-represented.
- Mismatch 2: people outside the district, or under 18, can fill in the form even though they are not in the target population.
- One clearly described mismatch, in context, is enough.
Full working
- target population: adult residents of Te Awa District
- sampling frame: people who see the council's Facebook page
- mismatch: non-Facebook residents excluded; non-residents can respond
Survey A found 64% support but Survey B found 47% support.
Explain why Survey A's result is likely to be biased, and suggest why it is higher than Survey B's.
Hint
Who chose to fill in Survey A? Who is most likely to bother filling in a form about a cycleway?
Answer
Survey A is a self-selected (volunteer) sample: people chose whether to respond. People with strong views, such as keen cyclists, and people who follow the council online are more likely to respond, and cycling groups may have shared the link. If supporters were more motivated to respond, the 64% overestimates support. Survey B used a random sample, so its 47% is less likely to be biased.
Step by step
- In Survey A nobody was chosen at random; people decided for themselves to respond. This is a self-selected sample.
- People who care most about a cycleway, especially cyclists who would use it, are more likely to take the time to respond.
- A cycling club could also share the link, bringing in many supporters.
- So supporters are likely over-represented, which would push the result above the true level of support. This fits Survey A (64%) being higher than Survey B (47%).
- Survey B chose people at random from the electoral roll, so it is less likely to be biased in this way.
Full working
- Survey A: self-selected, so people with strong views respond
- supporters (cyclists) more motivated; link may be shared by cycling groups
- supporters over-represented, so 64% likely too high; B is random, 47%
A student used the rule of thumb to give a 95% confidence interval for the percentage of adult residents who cycle to work at least once a week: 12% ± 4.1%.
Explain why the rule of thumb is not suitable here, and what effect it has on the interval.
Hint
The 1√n rule of thumb works well for percentages in a certain range. Is 12% in that range?
Answer
12% is outside the 30% to 70% range where the 1√n rule of thumb works well. For a percentage this far from 50%, the rule of thumb overestimates the margin of error, so the interval 7.9% to 16.1% is wider than it needs to be.
Step by step
- Rule of thumb: 1√600 = 0.0408, about 4.1%, so the interval is 12% ± 4.1% = 7.9% to 16.1%.
- The 1√n rule of thumb is based on a percentage of 50%, where the margin of error is largest. It works well for survey percentages between about 30% and 70%.
- 12% is outside that range, so the true margin of error is smaller than 4.1%.
- Using the rule of thumb therefore gives a confidence interval that is wider than it needs to be.
Full working
- 1√600 = 0.0408 = 4.1%; 12% ± 4.1% = [7.9%, 16.1%]
- 12% is outside 30% to 70%
- rule of thumb overestimates the MOE, so the interval is too wide
A councillor claims: "Survey B shows that a majority of the district's adult residents support the cycleway."
Use an appropriate confidence interval from Survey B to evaluate this claim, and write an inference statement in context.
Hint
Find 1√600, build the interval around 47%, then check where 50% sits.
Answer
MOE = 1√600 = 4.1%. CI 47% ± 4.1% = 42.9% to 51.1%. We are pretty sure that between 42.9% and 51.1% of the district's adult residents support the cycleway. The interval includes values below and above 50%, so we cannot say that a majority support it: the claim is not supported.
Step by step
- Survey B is a random sample of 600, so MOE = 1√600 = 0.0408, about 4.1%.
- Confidence interval: 47% ± 4.1% = 42.9% to 51.1%.
- Inference: we are pretty sure that the proportion of all adult residents of Te Awa District who support the cycleway is between 42.9% and 51.1%.
- A majority means more than 50%. The interval includes 50% and values on both sides of it.
- So the survey does not give enough evidence that a majority support the cycleway. It also does not show that a majority oppose it.
- Survey A's 64% cannot be used to back the claim: it is self-selected, so a confidence interval from it is not valid.
Full working
- MOE = 1√600 = 0.0408 = 4.1%
- CI = 47% ± 4.1% = [42.9%, 51.1%]
- interval contains 50%, so a majority is not shown; the claim is not supported
Evaluate the council's statement about Survey A: "With 2500 responses, our margin of error is only 2%."
Hint
Check the arithmetic first, then ask what a margin of error actually measures, and what kind of sample it needs.
Answer
The arithmetic is right: 1√2500 = 0.02 = 2%. But a margin of error only measures sampling error for a random sample. Survey A is self-selected, so it is likely biased, and a large sample does not remove bias: 2500 self-selected responses can still be far from the true level of support. The 2% gives a false impression of precision; the random Survey B (47% ± 4.1%) is the more trustworthy estimate even though its sample is smaller.
Step by step
- Arithmetic: 1√2500 = 150 = 0.02, so 2% is what the rule of thumb gives.
- But the margin of error describes only the variation from one random sample to another. It assumes the sample was chosen at random.
- Survey A was self-selected, so there is no random sampling behind it, and its margin of error has no real meaning.
- Bigger samples reduce sampling error, but they do not reduce bias. 2500 self-selected responses can be just as biased as 250.
- The large gap between Survey A (64%) and Survey B (47%) is far more than 2%, which suggests Survey A is biased.
- So the statement is misleading: Survey B, with an interval of 42.9% to 51.1%, is the better estimate of support.
Full working
- 1√2500 = 0.02 = 2%arithmetic correct
- MOE only measures sampling error for a random sample
- Survey A is self-selected: bias is not reduced by a large sample
- 64% vs 47% differ by far more than 2%; the statement is misleading
AS 91584 past papers · All NCEA Level 3 Maths
A Vertex Academy practice paper, written from NZQA's 2026 assessment specification for AS 91584. It is not an NZQA paper. Grades use the 2025 exam's cut scores: 8, 15 and 20 out of 24 for Achieved, Merit and Excellence.