Vertex Learn

Level 3 Probability concepts practice paper 2026

Question One

Question One

a
Achieved

A Queenstown tourism survey asked 1500 visitors which of three activities they did during their stay: a bungy jump, a jet boat ride, or the gondola.

• 420 did a bungy jump, 610 rode a jet boat and 880 rode the gondola.

• 240 did both a bungy jump and a jet boat ride.

• 300 did both a jet boat ride and the gondola.

• 220 did both a bungy jump and the gondola.

• 90 did all three.

Estimate the probability that a randomly selected visitor from the survey did exactly one of the three activities.

Hint

Draw a Venn diagram with three circles. Put the 90 in the centre first, then fill the 'two only' regions, then the 'one only' regions.

Answer

Bungy only 50, jet boat only 160, gondola only 450, so 660 visitors did exactly one activity. P(exactly one) = 6601500 = 0.44.

Step by step
  1. Draw three overlapping circles: B (bungy), J (jet boat) and G (gondola).
  2. Start in the centre: 90 did all three.
  3. Two only: bungy and jet boat only = 240 − 90 = 150. Jet boat and gondola only = 300 − 90 = 210. Bungy and gondola only = 220 − 90 = 130.
  4. One only: take everything else in that circle away from its total.
  5. Bungy only = 420 − 150 − 130 − 90 = 50.
  6. Jet boat only = 610 − 150 − 210 − 90 = 160.
  7. Gondola only = 880 − 210 − 130 − 90 = 450.
  8. Exactly one activity: 50 + 160 + 450 = 660 visitors.
  9. P(exactly one) = 6601500 = 0.44.
  10. Check: the eight regions add to 1500 (50 + 160 + 450 + 150 + 210 + 130 + 90, plus 260 who did none).
  11. Classic mistake: subtracting only the 'both' totals (420 − 240 − 220) and forgetting that the 90 has then been taken away twice.
Full working
  1. all three = 90centre of the Venn diagram
  2. two only: 150, 210, 130subtract 90 from each pair
  3. B only = 420 − 150 − 130 − 90 = 50
  4. J only = 610 − 150 − 210 − 90 = 160
  5. G only = 880 − 210 − 130 − 90 = 450
  6. exactly one = 50 + 160 + 450 = 660
  7. P(exactly one) = 6601500 = 0.44
b
Merit

A Queenstown tourism survey asked 1500 visitors which activities they did. 420 did a bungy jump, 610 rode a jet boat, and 240 did both a bungy jump and a jet boat ride.

Interpret the relationship between the events 'visitor did a bungy jump' and 'visitor rode a jet boat'. As part of your answer, explain whether these two events are independent, and whether they are mutually exclusive.

Use calculations and statistical reasoning to support your answer.

Hint

Compare P(bungy | jet boat) with P(bungy). If knowing one event changes the chance of the other, they are not independent. Mutually exclusive means they can never happen together.

Answer

P(B) = 4201500 = 0.28 but P(B | J) = 240610 = 0.3934. Visitors who rode a jet boat were more likely to bungy jump than visitors in general, so the events are not independent (also P(B) × P(J) = 0.1139 but P(B and J) = 0.16). They are not mutually exclusive because 240 visitors did both. In context, visitors who choose one thrill activity are more likely to choose another.

Step by step
  1. Let B = did a bungy jump and J = rode a jet boat.
  2. P(B) = 4201500 = 0.28. This is the chance for any visitor.
  3. P(B | J) means: of the jet boat riders only, what fraction did a bungy jump? Only the 610 jet boat riders count, so P(B | J) = 240610 = 0.3934.
  4. 0.3934 is bigger than 0.28. Knowing a visitor rode a jet boat raises the chance they bungy jumped, so B and J are not independent.
  5. The multiplication test says the same: P(B) × P(J) = 0.28 × 0.4067 = 0.1139, but P(B and J) = 2401500 = 0.16. These are not equal.
  6. Mutually exclusive means the events cannot happen together. 240 visitors did both, so P(B and J) = 0.16, not 0. They are not mutually exclusive.
  7. In context: there is a positive relationship. Visitors who like one adrenaline activity are more likely to do another one too.
  8. Classic mistake: mixing up the two ideas. Independent means one event does not change the chance of the other; mutually exclusive means they never happen together.
Full working
  1. P(B) = 4201500 = 0.28
  2. P(BJ) = 240610 = 0.3934 | only the 610 jet boat riders
  3. 0.3934 ≠ 0.28, so not independentalso P(B) × P(J) = 0.1139 ≠ 0.16
  4. P(B and J) = 0.16 ≠ 0, so not mutually exclusive
  5. Not independent (jet boat riders more likely to bungy), not mutually exclusive
c
Merit

A Queenstown tourism survey asked 1500 visitors which of three activities they did: a bungy jump (420 visitors), a jet boat ride (610) or the gondola (880). 240 did bungy and jet boat, 300 did jet boat and gondola, 220 did bungy and gondola, and 90 did all three.

(i) A visitor is randomly selected from those who did at least two of the activities. Find the probability that they did all three.

(ii) A tour operator claims that more than half of the visitors who did a bungy jump also rode the gondola. Does the survey support this claim?

Hint

(i) Count everyone in the overlapping regions of the Venn diagram: the three 'two only' regions plus the centre. That total is your denominator. (ii) The claim is about bungy jumpers only, so divide by 420.

Answer

(i) At least two: 150 + 210 + 130 + 90 = 580. P(all three | at least two) = 90580 = 0.1552. (ii) P(gondola | bungy) = 220420 = 0.5238, which is more than 0.5, so the survey supports the claim, though only just.

Step by step
  1. (i) The 'two only' regions are: bungy and jet boat only 240 − 90 = 150, jet boat and gondola only 300 − 90 = 210, bungy and gondola only 220 − 90 = 130.
  2. At least two activities means two only or all three: 150 + 210 + 130 + 90 = 580 visitors.
  3. Given 'at least two' means the 580 visitors are the whole group now. 90 of them did all three.
  4. P(all three | at least two) = 90580 = 0.1552.
  5. Classic mistake: adding 240 + 300 + 220 + 90. That counts the 90 four times.
  6. (ii) The claim is about visitors who did a bungy jump, so the denominator is 420.
  7. 220 bungy jumpers rode the gondola (this already includes the 90 who did all three).
  8. P(gondola | bungy) = 220420 = 0.5238, which is 52.4%.
  9. This is more than half, so the survey supports the claim.
  10. It is only just over 50%, and it is one survey of 1500 visitors (perhaps at one time of year), so another sample could fall below half.
Full working
  1. two only: 150 + 210 + 130 = 490subtract 90 from each pair
  2. at least two = 490 + 90 = 580
  3. P(all threeat least two) = 90580 = 0.1552
  4. P(GB) = 220420 = 0.5238 | bungy jumpers only
  5. 0.5238 > 0.5, so the claim is supported (only just)
d
Excellence

A Queenstown jet boat company takes bookings two ways. 35% of bookings are made online and the rest are made in person at an i-SITE visitor centre. Records show that 6% of online bookings are no-shows (the customer does not turn up), and 2% of i-SITE bookings are no-shows.

(i) Given that a booking was a no-show, calculate the probability that it was made online.

(ii) The manager says: 'Most of our no-shows are online bookings, even though most of our bookings are made at the i-SITE.' Use probabilities to explain how both parts of this statement can be true.

Hint

Draw a tree: first branch online or i-SITE, second branch no-show or turns up. P(no-show) is the sum of two branch products. For (ii), compare the risk of a no-show for each type of booking.

Answer

(i) P(no-show) = 0.35 × 0.06 + 0.65 × 0.02 = 0.034. P(online | no-show) = 0.0210.034 = 0.6176. (ii) Only 35% of bookings are online, but the no-show risk for online bookings is 0.060.02 = 3 times the i-SITE risk. That higher risk outweighs the smaller share, so 61.8% of no-shows are online bookings while 65% of all bookings are i-SITE ones.

Step by step
  1. (i) Draw a tree diagram. First branches: online 0.35, i-SITE 0.65.
  2. Second branches: online then no-show 0.06; i-SITE then no-show 0.02.
  3. Multiply along each branch: P(online and no-show) = 0.35 × 0.06 = 0.021. P(i-SITE and no-show) = 0.65 × 0.02 = 0.013.
  4. Add the no-show branches: P(no-show) = 0.021 + 0.013 = 0.034.
  5. Given a no-show, only those two branches count. P(online | no-show) = 0.0210.034 = 0.6176.
  6. Classic mistake: answering 0.06. That is P(no-show | online), the other way round.
  7. (ii) 'Most bookings are i-SITE' is true: 65% are.
  8. 'Most no-shows are online' is also true: from (i), 61.8% of no-shows are online.
  9. The reason is the risk. An online booking is a no-show with probability 0.06; an i-SITE booking with probability 0.02.
  10. Relative risk = 0.060.02 = 3. An online booking is three times as likely to be a no-show.
  11. So even though online bookings are the smaller group, each one is much more likely to be a no-show, and online bookings make up 0.021 out of the 0.034 no-show probability.
  12. In context: customers who book online have not met staff in person and may book several activities at once, so they may be more likely to drop one.
Full working
  1. P(online and no-show) = 0.35 × 0.06 = 0.021
  2. P(i-SITE and no-show) = 0.65 × 0.02 = 0.013
  3. P(no-show) = 0.021 + 0.013 = 0.034
  4. P(onlineno-show) = 0.0210.034 = 0.6176
  5. RR = 0.060.02 = 3online three times as likely to no-show
  6. 61.8% of no-shows are online although 65% of bookings are i-SITE: the higher risk outweighs the smaller share
e
Excellence

A Queenstown jet boat company's records show that 35% of bookings are made online and 65% at an i-SITE. 6% of online bookings and 2% of i-SITE bookings are no-shows, so overall 3.4% of bookings are no-shows.

The boat has 12 seats. The company is thinking of selling 13 tickets for each trip. It will only do this if the probability that someone is turned away (all 13 ticket holders turn up) is less than 0.5.

(i) Estimate the probability that all 13 ticket holders turn up, using the overall no-show rate. State the assumption you have made.

(ii) The company is also thinking of overbooking only trips sold entirely online. Find the probability that all 13 turn up for such a trip, and advise the company.

(iii) Many tickets are bought by families and groups travelling together. Discuss how this affects your answers and your advice.

Hint

If each ticket holder turns up independently with probability 1 − 0.034, multiply 13 of these together. For (iii), think about what happens to a group of four if one of them is sick.

Answer

(i) P(all 13 turn up) = 0.966¹³ = 0.6378, assuming each ticket holder turns up independently with the same probability. (ii) 0.94¹³ = 0.4474, which is under 0.5, so overbooking would meet the rule only on online-only trips; the rule still allows someone to be turned away on about 45% of trips. (iii) Group members turn up or miss together, so the 13 tickets act like fewer independent bookings. Then all 13 turn up more often (for example 4 online groups: 0.94⁴ = 0.7807), so the answers above are underestimates and overbooking is riskier than the calculation suggests.

Step by step
  1. (i) P(a ticket holder turns up) = 1 − 0.034 = 0.966.
  2. Someone is turned away only if all 13 turn up. If they act independently: P(all 13) = 0.966¹³ = 0.6378.
  3. Assumption: each ticket holder turns up independently of the others, with the same 0.966 chance (the overall booking mix of online and i-SITE).
  4. 0.6378 is more than 0.5, so mixed trips fail the company's rule.
  5. (ii) For online tickets, P(turns up) = 1 − 0.06 = 0.94. P(all 13) = 0.94¹³ = 0.4474.
  6. 0.4474 is less than 0.5, so online-only trips meet the rule.
  7. Advice: overbooking could be tried on online-only trips, but even then a customer is turned away on about 45 trips in every 100. That is a lot of unhappy customers and refunds, so the company should weigh the cost against one extra fare.
  8. (iii) The independence assumption is unlikely to hold. A family of four either all turn up or all miss together, for example if one is sick or their flight is late.
  9. Then the 13 tickets behave like a few large bookings, not 13 separate people. With fewer independent 'chances' of a no-show, it is more likely that everyone turns up.
  10. Example: 13 online tickets bought as 4 groups, each missing together with probability 0.06: P(all turn up) = 0.94⁴ = 0.7807, much higher than 0.4474.
  11. So the calculations in (i) and (ii) underestimate the chance of turning someone away. With groups, even online-only trips probably fail the 0.5 rule, and the company should not overbook without checking real records of how many ticket holders turn up per trip.
Full working
  1. P(all 13 turn up) = (1 − 0.034)¹³ = 0.966¹³ = 0.6378assume independent, same probability
  2. 0.6378 > 0.5mixed trips fail the rule
  3. online: 0.94¹³ = 0.4474 < 0.5meets the rule, but someone turned away on about 45% of trips
  4. groups act together, e.g. 4 groups: 0.94⁴ = 0.7807
  5. Independence fails for groups, so the risk is underestimated; overbooking is riskier than calculated
Question Two

Question Two

a
Achieved

Detector dogs at Auckland Airport sniff incoming bags for undeclared biosecurity risk goods, such as fruit or meat. If a dog sits beside a bag, it 'indicates' the bag. In a trial, 2000 bags were checked by a dog and then searched by hand. The results are in the table.

Find the probability that a randomly selected bag from the trial had risk goods or was indicated by the dog (or both).

Risk goodsNo risk goodsTotal
Dog indicates152216368
Dog does not indicate3815941632
Total19018102000
Hint

Add the three cells that are in the 'Risk goods' column or the 'Dog indicates' row. Do not count the 152 twice.

Answer

152 + 216 + 38 = 406 bags. P(risk goods or indicates) = 4062000 = 0.203.

Step by step
  1. 'Or' means at least one of the two things happened.
  2. Cells with risk goods or an indication: 152 (both), 216 (indicated, no risk goods) and 38 (risk goods, not indicated).
  3. Add them: 152 + 216 + 38 = 406.
  4. P(risk goods or indicates) = 4062000 = 0.203.
  5. Check with the complement: the only bags left out are the 1594 with neither, and 1 − 15942000 = 0.203.
  6. Or use the addition rule: 1902000 + 3682000 − 1522000 = 4062000.
  7. Classic mistake: adding 190 + 368 = 558. That counts the 152 bags in both groups twice.
Full working
  1. 152 + 216 + 38 = 406cells with risk goods or an indication
  2. check: 2000 − 1594 = 406
  3. P = 4062000 = 0.203
b
Merit

In a trial at Auckland Airport, 2000 bags were checked by a detector dog and then searched by hand for biosecurity risk goods. The results are in the table.

Two claims are made about the dog.

Claim A: 'The dog finds four out of every five bags that contain risk goods.'

Claim B: 'When the dog sits beside a bag, the bag usually contains risk goods.'

Does the trial data support each claim? Support your answers with statistical reasoning.

Risk goodsNo risk goodsTotal
Dog indicates152216368
Dog does not indicate3815941632
Total19018102000
Hint

Each claim is a conditional probability. Claim A starts from bags with risk goods (divide by 190). Claim B starts from bags the dog indicated (divide by 368).

Answer

Claim A: P(indicates | risk goods) = 152190 = 0.8, exactly four in five, so it is supported. Claim B: P(risk goods | indicates) = 152368 = 0.4130, so only about 41% of indicated bags have risk goods. Most (58.7%) do not, so Claim B is not supported.

Step by step
  1. Claim A is about bags that contain risk goods. There are 190 of them.
  2. The dog indicated 152 of these. P(indicates | risk goods) = 152190 = 0.8.
  3. Four out of five is 0.8, so the trial supports Claim A.
  4. Claim B is about bags the dog indicated. There are 368 of them.
  5. 152 of these had risk goods. P(risk goods | indicates) = 152368 = 0.4130.
  6. 'Usually' means more than half. 0.4130 is less than 0.5, so Claim B is not supported. In fact 216368 = 0.5870 of indicated bags had no risk goods.
  7. Why the difference? Most bags (1810 of 2000) have no risk goods, so even a small rate of wrong indications on those bags gives more false alarms (216) than correct finds (152).
  8. Classic mistake: treating P(indicates | risk goods) and P(risk goods | indicates) as the same thing. They answer different questions.
Full working
  1. P(indicatesrisk goods) = 152190 = 0.8 | Claim A supported
  2. P(risk goodsindicates) = 152368 = 0.4130
  3. 0.4130 < 0.5Claim B not supported
  4. A supported; B not supported (58.7% of indicated bags had no risk goods)
c
Merit

A study followed 840 Auckland adults who cycle to work for one year. 360 of them rode mostly on separated bike lanes, and 18 of these were injured in a crash during the year. The other 480 rode mostly on roads with traffic, and 48 of these were injured.

(i) Calculate the relative risk of injury for road riders compared with bike-lane riders, and interpret it in context.

(ii) Give one reason why this study cannot show that riding on roads causes the extra injuries.

Hint

Risk for a group = injured in that group ÷ size of that group. Relative risk = road risk ÷ bike-lane risk.

Answer

(i) Risk for road riders = 48480 = 0.10; risk for bike-lane riders = 18360 = 0.05. Relative risk = 0.100.05 = 2. Road riders in this study were twice as likely to be injured as bike-lane riders. (ii) It is an observational study: riders chose their route, so the groups may differ in other ways, such as distance ridden, experience, speed or the type of bike.

Step by step
  1. (i) Risk is a conditional probability: the chance of injury given the group.
  2. Road riders: P(injured | road) = 48480 = 0.10.
  3. Bike-lane riders: P(injured | bike lane) = 18360 = 0.05.
  4. Relative risk (road compared with bike lane) = 0.100.05 = 2.
  5. Interpretation: in this study, road riders were 2 times as likely (twice as likely) to be injured in a crash during the year as bike-lane riders.
  6. Classic mistake: dividing by 840 for each group. Each risk uses its own group total.
  7. (ii) The riders were not randomly put into the groups. They chose how they ride.
  8. So the groups could differ in ways that also affect injuries. For example, road riders may ride further each day, ride faster, or ride in busier parts of the city.
  9. Any of these could cause the extra injuries, so we cannot say the road itself is the cause.
Full working
  1. risk(road) = 48480 = 0.10
  2. risk(lane) = 18360 = 0.05
  3. RR = 0.100.05 = 2
  4. Road riders twice as likely to be injured; observational, so other differences (e.g. distance ridden) could explain it
d
Excellence

A study followed 840 Auckland adults who cycle to work for one year. Of 360 who rode mostly on separated bike lanes, 18 were injured in a crash. Of 480 who rode mostly on roads with traffic, 48 were injured. The study also recorded how far each person rode: the bike-lane riders averaged 45 km a week and the road riders averaged 90 km a week.

A news website reports: 'Riding on the road doubles your risk of injury. Moving all 480 road riders onto bike lanes would prevent 24 injuries a year.'

Evaluate this report. Use calculations and statistical reasoning to support your answer.

Hint

First find the relative risk per rider. Then find the number of injuries per kilometre ridden for each group, using total weekly kilometres (riders × average km). What does the comparison per kilometre show?

Answer

Per rider, the risks are 48480 = 0.10 and 18360 = 0.05, a relative risk of 2, and 480 × 0.05 = 24 is where '24 fewer injuries' comes from. But road riders rode twice as far. Injuries per 1000 km of weekly riding: road 4843 200 × 1000 = 1.11, bike lane 1816 200 × 1000 = 1.11, exactly the same. Per kilometre ridden there is no difference in risk, so the higher risk per rider is explained by distance, not the road. The report is not supported: the study is observational, the '24 injuries prevented' figure assumes the road causes the difference, and road riders moved to bike lanes would still ride 90 km a week.

Step by step
  1. Check the report's numbers first. Risk(road) = 48480 = 0.10; risk(bike lane) = 18360 = 0.05. Relative risk = 2, so 'doubles' matches the raw data per rider.
  2. The 24 comes from: if the 480 road riders had the bike-lane risk of 0.05, there would be 480 × 0.05 = 24 injuries instead of 48, so 24 fewer.
  3. Now use the distances. Total weekly km for bike-lane riders: 360 × 45 = 16 200 km. For road riders: 480 × 90 = 43 200 km.
  4. Injuries per 1000 km of weekly riding: bike lane 1816 200 × 1000 = 1.11; road 4843 200 × 1000 = 1.11.
  5. The risk per kilometre is the same for both groups. The road riders were injured twice as often because they rode twice as far, not because of where they rode.
  6. So distance is a confounding variable: it is linked to both the type of route and the number of injuries.
  7. The '24 injuries prevented' claim assumes the road causes the extra risk. If road riders switched to bike lanes but still rode 90 km a week, this study gives no evidence their injuries would drop.
  8. Other issues: this is one observational study in one city over one year, the injury counts are small (18 and 48), and other differences such as speed, e-bikes or experience were not measured.
  9. Conclusion: the headline is not supported by the study. A fair statement is that road riders had twice the injury rate per person, but the same rate per kilometre ridden.
Full working
  1. risk(road) = 48480 = 0.10, risk(lane) = 18360 = 0.05, RR = 2
  2. 480 × 0.05 = 24source of '24 injuries prevented'
  3. weekly km: lane 360 × 45 = 16 200, road 480 × 90 = 43 200
  4. per 1000 weekly km: lane 1816 200 × 1000 = 1.11, road 4843 200 × 1000 = 1.11
  5. Same risk per km: distance explains the RR of 2, so the report is not supported (observational, causation assumed)
e
Excellence

In a trial at Auckland Airport, a detector dog correctly indicated 80% of bags that contained biosecurity risk goods. It also wrongly indicated 216 of the 1810 bags that had no risk goods.

The same dog is moved to an arrivals hall where only 2% of bags contain risk goods.

(i) Assuming the dog performs as it did in the trial, find the probability that a bag the dog indicates in the new hall contains risk goods.

(ii) Biosecurity staff decide that every bag the dog indicates is also X-rayed, and the bag is only opened if the X-ray also flags it. The X-ray flags 90% of bags with risk goods and 5% of bags without. Find the probability that an opened bag contains risk goods, and the probability that a bag with risk goods is not opened. State the assumption you need, and evaluate this two-step system.

Hint

False indication rate = 2161810. Use a tree with first branch risk goods 0.02 or not 0.98, then dog, then X-ray. Multiply along branches and divide by the total for 'flagged'.

Answer

(i) False indication rate = 2161810 = 0.1193. P(indicates) = 0.02 × 0.8 + 0.98 × 0.1193 = 0.1330. P(risk goods | indicates) = 0.0160.1330 = 0.1203. (ii) P(risk and both flag) = 0.02 × 0.8 × 0.9 = 0.0144; P(no risk and both flag) = 0.98 × 0.1193 × 0.05 = 0.005848. P(risk goods | opened) = 0.01440.020248 = 0.7112. P(not opened | risk goods) = 1 − 0.8 × 0.9 = 0.28. This assumes the X-ray result is independent of the dog's result for a given bag. The second step makes opened bags much more likely to hold risk goods (71% against 12%), but 28% of risk bags now get through, compared with 20% using the dog alone.

Step by step
  1. (i) The dog's false indication rate is 2161810 = 0.1193 (it indicates about 12% of clean bags).
  2. Tree: risk goods 0.02, then indicates 0.8: 0.02 × 0.8 = 0.016.
  3. No risk goods 0.98, then indicates 0.1193: 0.98 × 0.1193 = 0.1170.
  4. P(indicates) = 0.016 + 0.1170 = 0.1330.
  5. P(risk goods | indicates) = 0.0160.1330 = 0.1203.
  6. This is much lower than the 0.4130 in the trial, because risk goods are rarer in the new hall, so false alarms from clean bags swamp the correct finds.
  7. (ii) Add an X-ray branch after 'indicates'. Risk goods: 0.02 × 0.8 × 0.9 = 0.0144. No risk goods: 0.98 × 0.1193 × 0.05 = 0.005848.
  8. P(opened) = 0.0144 + 0.005848 = 0.020248.
  9. P(risk goods | opened) = 0.01440.020248 = 0.7112.
  10. A bag with risk goods is opened only if the dog AND the X-ray flag it: 0.8 × 0.9 = 0.72. So P(not opened | risk goods) = 1 − 0.72 = 0.28.
  11. Assumption: for a given bag, the X-ray result does not depend on whether the dog indicated it. This may not hold: a bag packed in a way that hides food from the dog, such as sealed containers, might also be harder to spot on the X-ray, so more risk bags would get through.
  12. Evaluation: the two-step system wastes far less staff time, since 71% of opened bags now hold risk goods instead of 12%. But it lets 28% of risk bags through, compared with 20% using the dog alone. For biosecurity, a missed bag of fruit could bring in a pest such as fruit fly, so staff need to decide whether the time saved is worth the extra risk.
Full working
  1. false indication rate = 2161810 = 0.1193
  2. P(indicates) = 0.02 × 0.8 + 0.98 × 0.1193 = 0.1330
  3. P(riskindicates) = 0.0160.1330 = 0.1203
  4. P(risk and both flag) = 0.0144, P(no risk and both flag) = 0.0058assume dog and X-ray independent for a given bag
  5. P(riskopened) = 0.01440.020248 = 0.7112
  6. P(not openedrisk) = 1 − 0.72 = 0.28: fewer wasted searches, but more risk bags missed
Question Three

Question Three

a
Achieved

A Hamilton driving school recorded how many attempts its learners needed to pass the practical test for a restricted licence. The random variable X is the number of attempts a learner needs. Its probability distribution is shown in the table.

Find the value of k, and the probability that a learner needs three or more attempts.

X (attempts)12345 or more
P(X = x)0.420.280.150.08k
Hint

All the probabilities in a distribution add to 1. Three or more attempts means X = 3, 4, or 5 or more.

Answer

k = 1 − (0.42 + 0.28 + 0.15 + 0.08) = 0.07. P(X ≥ 3) = 0.15 + 0.08 + 0.07 = 0.30.

Step by step
  1. Every learner needs some number of attempts, so the probabilities add to 1.
  2. 0.42 + 0.28 + 0.15 + 0.08 = 0.93.
  3. k = 1 − 0.93 = 0.07.
  4. Three or more attempts: P(X ≥ 3) = P(X = 3) + P(X = 4) + P(X ≥ 5) = 0.15 + 0.08 + 0.07 = 0.30.
  5. Check with the complement: P(X ≥ 3) = 1 − P(X ≤ 2) = 1 − (0.42 + 0.28) = 0.30.
  6. In context: about 30% of the school's learners need three or more attempts.
Full working
  1. 0.42 + 0.28 + 0.15 + 0.08 = 0.93
  2. k = 1 − 0.93 = 0.07probabilities add to 1
  3. P(X ≥ 3) = 0.15 + 0.08 + 0.07
  4. k = 0.07, P(X ≥ 3) = 0.30
b
Merit

A Hamilton driving school recorded how many attempts its learners needed to pass the practical test for a restricted licence. X is the number of attempts. The distribution is in the table.

The school makes two claims in an advertisement.

Claim 1: 'Twice as many of our learners pass first time as need three or more attempts.'

Claim 2: 'If you fail your first attempt, you are more likely than not to pass on your second.'

Does the distribution support each claim? Use probabilities to justify your answers.

X (attempts)12345 or more
P(X = x)0.420.280.150.080.07
Hint

Claim 1: compare P(X = 1) with P(X ≥ 3). Claim 2: failing the first attempt means X ≥ 2, so you want P(X = 2 | X ≥ 2).

Answer

Claim 1: P(X = 1) = 0.42 and P(X ≥ 3) = 0.30. 0.420.30 = 1.4, not 2 (twice would need 0.60), so it is not supported. Claim 2: P(X = 2 | X ≥ 2) = 0.280.58 = 0.4828, which is less than 0.5, so it is not supported (just under half pass on their second attempt).

Step by step
  1. Claim 1: P(pass first time) = P(X = 1) = 0.42.
  2. P(three or more attempts) = 0.15 + 0.08 + 0.07 = 0.30.
  3. Twice 0.30 would be 0.60. The ratio is 0.420.30 = 1.4, so 1.4 times as many pass first time. Claim 1 is not supported.
  4. Claim 2 is a conditional probability. 'Fail your first attempt' means the learner needs 2 or more attempts.
  5. P(X ≥ 2) = 1 − 0.42 = 0.58. This is the new denominator.
  6. P(X = 2 | X ≥ 2) = 0.280.58 = 0.4828.
  7. 'More likely than not' means more than 0.5. 0.4828 is less than 0.5, so Claim 2 is not supported, though it is close.
  8. Classic mistake: using P(X = 2) = 0.28 for Claim 2. That ignores the information that the learner has already failed once.
Full working
  1. P(X = 1) = 0.42, P(X ≥ 3) = 0.30
  2. 0.420.30 = 1.4 ≠ 2Claim 1 not supported
  3. P(X ≥ 2) = 1 − 0.42 = 0.58
  4. P(X = 2X ≥ 2) = 0.280.58 = 0.4828
  5. 0.4828 < 0.5Claim 2 not supported
c
Merit

A cereal company puts one of six native bird cards (kākāpō, tūī, kea, kererū, pīwakawaka and takahē) in each box. A student assumes each card is equally likely and simulates collecting the full set by rolling a fair dice, counting rolls until all six numbers have appeared. One trial gives the number of boxes needed. The results of 40 trials are in the table.

(i) Use the simulation to estimate the probability that a family needs more than 15 boxes to collect the full set.

(ii) The company claims that most families will collect all six cards within 12 boxes. Does the simulation support this claim? Comment on how reliable your conclusion is.

Boxes needed6 to 910 to 1213 to 1516 to 1819 to 2122 to 2425 or more
Number of trials61097422
Hint

More than 15 boxes means the groups from 16 upwards. Within 12 boxes means the first two groups. 'Most' means more than half.

Answer

(i) 7 + 4 + 2 + 2 = 15 trials, so P(more than 15 boxes) ≈ 1540 = 0.375. (ii) 6 + 10 = 16 trials needed 12 or fewer, so P(within 12) ≈ 1640 = 0.4. This is less than half, so the simulation does not support the claim. With only 40 trials the estimate could easily vary by about 0.16 (1√40), so more trials (hundreds) would give a more reliable estimate, but 0.4 is far enough below 0.5 to make the claim doubtful.

Step by step
  1. (i) More than 15 boxes: the groups 16 to 18, 19 to 21, 22 to 24 and 25 or more.
  2. 7 + 4 + 2 + 2 = 15 trials.
  3. Experimental estimate: P(more than 15 boxes) ≈ 1540 = 0.375.
  4. (ii) Within 12 boxes: the groups 6 to 9 and 10 to 12. 6 + 10 = 16 trials.
  5. P(within 12 boxes) ≈ 1640 = 0.4.
  6. 'Most families' would need more than 0.5. The simulation gives 0.4, so it does not support the claim.
  7. Reliability: 40 trials is a small number. A rough margin of error is 1√40 ≈ 0.16, so the real value could be anywhere from about 0.24 to 0.56.
  8. More trials, say 500 or 1000 using a computer, would give a more precise estimate.
  9. The simulation also rests on the assumption that all six cards are equally likely. If they are not, the results could be quite different.
Full working
  1. more than 15: 7 + 4 + 2 + 2 = 15 trials
  2. P(more than 15) ≈ 1540 = 0.375
  3. within 12: 6 + 10 = 16 trials, P ≈ 1640 = 0.4
  4. 0.4 < 0.5claim not supported
  5. Not supported, but 40 trials is few (margin about ±0.16); run more trials
d
Excellence

A cereal company puts one of six native bird cards in each box. A student simulated collecting the cards assuming all six are equally likely, so P(kākāpō card) = 16.

The company later reveals that it printed 90 000 cards for the promotion, of which 9000 were kākāpō cards. A Year 13 class opened 75 boxes bought from local shops and found 6 kākāpō cards.

Compare the true probability, the model estimate and the experimental estimate of getting a kākāpō card in a box. Discuss which is the most reliable, and what this means for the student's simulation. The student's simulation estimated that 40% of families would collect all six cards within 12 boxes, and the company claims that most families will collect all six within 12 boxes.

Hint

True: from the whole population of cards printed. Model: from the equally likely assumption. Experimental: from the class's 75 boxes. Then think about how a rarer kākāpō card changes the number of boxes needed for a full set.

Answer

True probability = 900090 000 = 0.1. Model estimate = 16 = 0.1667. Experimental estimate = 675 = 0.08. The model is too high because the cards are not equally likely. The experimental estimate is close to the true value but based on only 75 boxes from local shops, so it is less precise. The true probability, from all cards printed, is the most reliable (if cards were packed randomly). Because the kākāpō card is rarer than the simulation assumed, families will need more boxes than the simulation suggests, so the simulation overestimates the chance of a full set within 12 boxes, and the company's claim is even less likely to be true.

Step by step
  1. True probability: this uses the whole population, every card printed. 900090 000 = 0.1.
  2. Model estimate: from the assumption that the six cards are equally likely. 16 = 0.1667.
  3. Experimental estimate: from the class's trial of 75 boxes. 675 = 0.08.
  4. Compare: the model (0.1667) is well above the true value (0.1). The experimental estimate (0.08) is below it but much closer.
  5. The model is wrong because its assumption is wrong: the company printed fewer kākāpō cards than the others.
  6. The experimental estimate is reasonable but uses only 75 boxes, so chance variation is large (a margin of error of about 1√75 ≈ 0.12). The boxes also came from local shops, which might hold boxes from the same batch.
  7. The true probability is the most reliable, because it is based on all 90 000 cards. It applies to a box only if the cards were mixed randomly into boxes across the country.
  8. Effect on the simulation: it used P(kākāpō) = 16. With the real 0.1, the kākāpō card is usually the last one collected and takes longer to find.
  9. So the simulation underestimates the number of boxes needed, and its estimate of 0.4 for a full set within 12 boxes is too high. The company's claim that most families finish within 12 boxes is even less likely to be true.
Full working
  1. true = 900090 000 = 0.1whole population of cards
  2. model = 16 = 0.1667equally likely assumption
  3. experimental = 675 = 0.08class sample
  4. model too high (assumption wrong); experimental close but small sample
  5. True probability most reliable; simulation underestimates boxes needed, so the claim is even less likely
e
Excellence

A cereal company printed 90 000 native bird cards: 9000 kākāpō cards and 16 200 each of the other five birds (tūī, kea, kererū, pīwakawaka and takahē). One card goes in each box.

(i) Describe how to carry out ONE trial of a simulation, using random whole numbers from 1 to 100, to estimate the number of boxes needed to collect all six cards with these probabilities.

(ii) The company also claims: 'Buy one box a week for 12 weeks and you are almost certain (at least 95%) to get a kākāpō card.' Evaluate this claim using the true probability, and compare with the answer you would get if all cards were equally likely. Find the least number of boxes needed to give at least a 95% chance of a kākāpō card. State the assumption you have made.

Hint

(i) P(kākāpō) = 900090 000 = 0.1 and each other bird is 16 20090 000 = 0.18. Give each card a block of numbers of the right size. (ii) P(at least one kākāpō in n boxes) = 1 − P(none) = 1 − 0.9ⁿ. Set this ≥ 0.95 and use logs.

Answer

(i) P(kākāpō) = 0.1 and each other bird 0.18. Assign 1 to 10 = kākāpō, 11 to 28 = tūī, 29 to 46 = kea, 47 to 64 = kererū, 65 to 82 = pīwakawaka, 83 to 100 = takahē. Generate random numbers, recording each bird, until all six have appeared; the count of numbers generated is the result of one trial. Repeat many times. (ii) True: P(at least one kākāpō in 12) = 1 − 0.9¹² = 0.7176, well below 0.95, so the claim is not supported. Equally likely model: 1 − (56)¹² = 0.8878, still below 0.95. Need 0.9ⁿ ≤ 0.05, so n ≥ 28.43: 29 boxes (17 under the equally likely model). Assumes boxes are independent, with cards randomly mixed.

Step by step
  1. (i) P(kākāpō) = 900090 000 = 0.1. Each other bird: 16 20090 000 = 0.18.
  2. Out of 100 numbers, give kākāpō 10 numbers and each other bird 18: 1 to 10 kākāpō, 11 to 28 tūī, 29 to 46 kea, 47 to 64 kererū, 65 to 82 pīwakawaka, 83 to 100 takahē.
  3. One trial: generate random numbers from 1 to 100 one at a time. Each number is one box. Note which bird it gives.
  4. Stop when all six birds have appeared. Record how many numbers (boxes) it took. That is the result of one trial.
  5. Repeat for many trials (at least several hundred) and use the results to estimate probabilities, such as P(full set within 12 boxes).
  6. (ii) P(no kākāpō in one box) = 0.9. Assuming boxes are independent, P(no kākāpō in 12 boxes) = 0.9¹² = 0.2824.
  7. P(at least one kākāpō in 12 boxes) = 1 − 0.2824 = 0.7176. That is about 72%, well below 95%, so the claim is not supported.
  8. If all cards were equally likely: 1 − (56)¹² = 1 − 0.1122 = 0.8878. Even under the model, 89% is still below 95%.
  9. Least number of boxes: need 1 − 0.9ⁿ ≥ 0.95, so 0.9ⁿ ≤ 0.05.
  10. Take logs: n log 0.9 ≤ log 0.05. log 0.9 is negative, so dividing by it flips the sign: n ≥ log 0.05 ÷ log 0.9 = 28.43.
  11. So 29 boxes are needed. Check: 1 − 0.9²⁸ = 0.9477 (not enough), 1 − 0.9²⁹ = 0.9529 (enough). Under the equally likely model only 17 boxes would be needed.
  12. Assumption: each box is independent, with P(kākāpō) = 0.1, which needs the cards to be mixed randomly through all boxes. If boxes in one shop come from the same batch, a family buying from that shop could get many boxes in a row without a kākāpō card, making the true chance even lower.
Full working
  1. P(kākāpō) = 0.1, others 0.18 each1 to 10 kākāpō, then blocks of 18
  2. one trial: generate numbers until all six birds appear; record the count
  3. P(at least one in 12) = 1 − 0.9¹² = 0.7176assume independent boxes
  4. model: 1 − (56)¹² = 0.8878
  5. 0.9ⁿ ≤ 0.05, n ≥ log 0.05 ÷ log 0.9 = 28.43log 0.9 < 0, sign flips
  6. Claim not supported (0.7176 < 0.95); 29 boxes needed

AS 91585 past papers · All NCEA Level 3 Maths

A Vertex Academy practice paper, written from NZQA's 2026 assessment specification for AS 91585. It is not an NZQA paper. Grades use the 2025 exam's cut scores: 7, 14 and 19 out of 24 for Achieved, Merit and Excellence.