Vertex Learn

Level 3 Probability distributions practice paper 2026

Question One

Question One

a
Achieved

The crossing time of a passenger ferry from downtown Auckland to Waiheke Island is modelled by a normal distribution with mean 42 minutes and standard deviation 3.5 minutes. Use this model to estimate the probability that a crossing takes more than 48 minutes.

Hint

Let X be the crossing time. Find P(X > 48) with the normal distribution, μ = 42 and σ = 3.5.

Answer

P(X > 48) = 0.0432

Step by step
  1. Say what is random first: X = the time a crossing takes, in minutes.
  2. The model is normal with μ = 42 and σ = 3.5, so X ~ N(42, 3.5²).
  3. 48 minutes is above the mean, so we want the right tail: P(X > 48).
  4. Standardise: z = 48 − 423.5 = 1.714. So 48 minutes is about 1.7 standard deviations above the mean.
  5. Calculator, normal cdf from 48 to a very large number with μ = 42, σ = 3.5: 0.0432.
  6. Check it is sensible: about 4% of crossings take more than 48 minutes. That is small, which fits a value well above the mean.
Full working
  1. X ~ N(42, 3.5²)X = crossing time in minutes
  2. z = 48 − 423.5 = 1.714
  3. P(X > 48) = P(Z > 1.714)right tail
  4. P(X > 48) = 0.0432
b
Merit

A commuter catches the 7.00 am ferry to Waiheke and the 8.00 am ferry back on the same morning.

(i) Use the model in part (a) to estimate the probability that exactly one of the two crossings takes more than 45 minutes.

(ii) State the assumption you made in part (i). Explain, in context, why it may not be valid and how this could affect your estimate.

Hint

Find p = P(X > 45) first. Exactly one long crossing can happen in two orders: long then normal, or normal then long.

Answer

(i) P(X > 45) = 0.1957, so P(exactly one) = 2 × 0.1957 × 0.8043 = 0.3148. (ii) Independence: the length of one crossing does not affect the other. Both crossings are on the same morning, so rough weather or a slow boat would make both long. Then the crossings are more likely to be both long or both short, and the true probability of exactly one long crossing is likely lower than 0.3148.

Step by step
  1. (i) First find the chance that one crossing takes more than 45 minutes.
  2. z = 45 − 423.5 = 0.857, and P(X > 45) = 0.1957.
  3. So the chance a crossing is not that long is 1 − 0.1957 = 0.8043.
  4. Exactly one long crossing happens two ways: the morning one is long and the return is not, or the other way round.
  5. Each way has probability 0.1957 × 0.8043 = 0.1574. Two ways: 2 × 0.1574 = 0.3148.
  6. (ii) Multiplying the probabilities assumes the two crossings are independent: how long one takes does not change how long the other takes.
  7. In context this is doubtful. Both crossings are on the same morning, so they share the same wind, swell and tide, and often the same boat and crew.
  8. If the weather is rough, both crossings are likely to be slow. So the times tend to go together.
  9. When times go together, 'both long' and 'both short' become more likely, so 'exactly one long' becomes less likely than 0.3148.
Full working
  1. P(X > 45) = P(Z > 0.857) = 0.1957X ~ N(42, 3.5²)
  2. P(X ≤ 45) = 1 − 0.1957 = 0.8043
  3. P(exactly one) = 2 × 0.1957 × 0.8043two orders, assuming independence
  4. P(exactly one) = 0.3148
  5. Independence doubtful: same morning, same weather and boat, so the estimate is likely too high
c
Merit

Ferries leave Matiatia wharf every 30 minutes. A visitor arrives at the wharf at a random time without checking the timetable. Their waiting time, W minutes, is modelled by a uniform distribution from 0 to 30 minutes.

(i) Describe the graph of this distribution, including its height, and find the mean and standard deviation of W.

(ii) The visitor has already waited 10 minutes. Find the probability that their total wait is more than 25 minutes.

Hint

A uniform graph is a flat rectangle with area 1. For part (ii), you already know W > 10, so divide by P(W > 10).

Answer

(i) A flat rectangle from 0 to 30 with height 130. Mean 15 minutes, SD = 30√12 = 8.66 minutes. (ii) P(W > 25 | W > 10) = (530)/(2030) = 0.25.

Step by step
  1. (i) Every waiting time from 0 to 30 minutes is equally likely, so the graph is a flat rectangle from 0 to 30.
  2. The total area must be 1. The width is 30, so the height is 130 ≈ 0.0333.
  3. The mean is halfway: 0 + 302 = 15 minutes.
  4. Standard deviation of a uniform distribution: b − a√12 = 30√12 = 8.66 minutes.
  5. (ii) This is a conditional probability. We know W > 10, so we only look at the part of the rectangle from 10 to 30.
  6. P(W > 10) = (30 − 10) × 130 = 2030.
  7. A total wait of more than 25 minutes means W > 25. Since 25 > 10, both conditions together are just W > 25.
  8. P(W > 25) = (30 − 25) × 130 = 530.
  9. P(W > 25 | W > 10) = (530)/(2030) = 520 = 0.25.
  10. Classic mistake: answering 530 = 0.167. That ignores the fact the visitor has already waited 10 minutes.
Full working
  1. W ~ uniform, a = 0, b = 30height 130
  2. E(W) = 15, SD(W) = 30√12 = 8.66 minutes
  3. P(W > 25W > 10) = P(W > 25)P(W > 10) | conditional probability
  4. = (530)/(2030) = 520
  5. P(W > 25W > 10) = 0.25
d
Excellence

A new operator takes over the Waiheke route with faster boats. In its first month it makes 400 crossings, and reports that 10% of them take longer than 47 minutes and 2% take less than 36 minutes. Assume crossing times are normally distributed.

(i) Find the mean and standard deviation of the new crossing times.

(ii) The operator advertises a '45 minute crossing'. Estimate the proportion of crossings that take longer than 45 minutes, and comment on the advert.

(iii) The operator wants only 5% of crossings to take longer than 45 minutes, without changing the standard deviation. Find the mean crossing time needed.

Hint

Use inverse normal on the standard normal to get two z-values, then write X = μ + zσ twice and solve the two equations.

Answer

(i) μ = 42.77 minutes, σ = 3.30 minutes. (ii) 0.2498: about a quarter of crossings take longer than the advertised 45 minutes, so the advert is misleading. (iii) μ = 39.58 minutes, about 3.2 minutes faster on average.

Step by step
  1. (i) 10% take longer than 47, so 90% take less than 47. Inverse normal with area 0.9: z = 1.2816.
  2. 2% take less than 36. Inverse normal with area 0.02: z = −2.0537.
  3. A value is the mean plus z standard deviations, so: μ + 1.2816σ = 47 and μ − 2.0537σ = 36.
  4. Subtract the second equation from the first: 3.3353σ = 11, so σ = 3.298 minutes.
  5. Then μ = 47 − 1.2816 × 3.298 = 42.77 minutes. Keep the unrounded values for the next parts.
  6. (ii) P(X > 45) with μ = 42.77 and σ = 3.298: z = 45 − 42.773.298 = 0.675, so the probability is 0.2498.
  7. About a quarter of crossings take longer than the advertised 45 minutes, so the advert is optimistic.
  8. (iii) For only 5% above 45, 45 must be at the 95th percentile: z = 1.6449.
  9. 45 = μ + 1.6449 × 3.298, so μ = 45 − 5.425 = 39.58 minutes. The average crossing must be about 3.2 minutes faster.
  10. Check: with μ = 39.58 and σ = 3.298, P(X > 45) = 0.05.
Full working
  1. μ + 1.2816σ = 47P(X < 47) = 0.9
  2. μ − 2.0537σ = 36P(X < 36) = 0.02
  3. 3.3353σ = 11, so σ = 3.298subtract
  4. μ = 47 − 1.2816 × 3.298 = 42.77
  5. P(X > 45) = P(Z > 0.675) = 0.2498a quarter, so the advert overstates
  6. 45 = μ + 1.6449 × 3.29895th percentile
  7. μ = 42.77, σ = 3.30; P(X > 45) = 0.2498; new mean 39.58 minutes
e
Excellence

Of the operator's 400 crossings in part (d), 31 took longer than 48 minutes, and the longest took 58 minutes.

Use the normal model from part (d) to evaluate whether a normal distribution is a suitable model for these crossing times. Discuss how your conclusion affects the mean found in part (d)(iii).

Hint

Find how many of 400 crossings the model expects to take longer than 48 minutes, and how unlikely a 58-minute crossing is under the model. Then think about the shape of real crossing times.

Answer

The model gives P(X > 48) = 0.0565, so about 22.6 of 400 crossings, but 31 happened. A 58-minute crossing is 4.6 standard deviations above the mean: P(X > 58) ≈ 0.000002, so it should almost never happen in 400 crossings. The upper tail is longer than the normal model allows, so crossing times are likely skewed to the right (a hard minimum set by top speed, but delays from weather and berthing can add many minutes). The normal model underestimates long crossings, so a mean of 39.58 minutes would still give more than 5% over 45 minutes; the operator would need a faster average or less variation.

Step by step
  1. Use the model X ~ N(42.77, 3.298²) from part (d).
  2. P(X > 48): z = 48 − 42.773.298 = 1.585, so P(X > 48) = 0.0565.
  3. Expected number of 400 crossings: 400 × 0.0565 = 22.6. The operator had 31, about 37% more than the model expects.
  4. The longest crossing: z = 58 − 42.773.298 = 4.62. P(X > 58) ≈ 0.000002, so in 400 crossings the model expects about 0.0008 crossings this long. It is very unlikely under the model.
  5. Both facts say the same thing: the real upper tail is longer than the normal model allows.
  6. In context this makes sense. A crossing cannot be much faster than the boat's top speed allows, but rough weather, a slow boat or waiting to berth can add a lot of time. So crossing times are likely skewed to the right, and a symmetric normal model does not fit the upper tail well.
  7. Effect on part (d)(iii): the mean of 39.58 minutes was found using the normal model's upper tail. If the real tail is longer, more than 5% of crossings would still take longer than 45 minutes.
  8. So the operator would need an even lower mean, or to reduce the variation (for example by cutting berthing delays). A right-skewed model fitted to the data would give a better answer.
  9. One month of data may also miss winter weather, which would make long crossings more common again.
Full working
  1. z = 48 − 42.773.298 = 1.585, P(X > 48) = 0.0565model from (d)
  2. expected 400 × 0.0565 = 22.6, observed 31
  3. z = 58 − 42.773.298 = 4.62, P(X > 58) ≈ 0.000002expected 0.0008 crossings
  4. observed upper tail longer than the model: right skew (lower limit, delays add time)
  5. Normal model not suitable for the upper tail; μ = 39.58 would give more than 5% over 45, so a lower mean or less variation is needed
Question Two

Question Two

a
Achieved

A vehicle testing station in Christchurch recorded how long 150 customers waited for a Warrant of Fitness (WOF) inspection. The shortest wait was 5 minutes and the longest was 64 minutes. The results are in the table.

The station's manager claims that at least one customer in five waits more than half an hour. Does this data support the claim?

Waiting time (minutes)Number of customers
5 to under 1014
10 to under 1533
15 to under 2030
20 to under 2524
25 to under 3017
30 to under 3512
35 to under 408
40 to under 455
45 to under 503
50 to under 552
55 to under 601
60 to under 651
Hint

Add up the customers in every row from 30 minutes upwards, then divide by 150. One in five is 0.2.

Answer

32 of the 150 customers waited 30 minutes or more: 32150 = 0.213, or 21.3%. This is more than one in five (20%), so the data supports the claim.

Step by step
  1. More than half an hour means the rows from 30 minutes upwards.
  2. Add them: 12 + 8 + 5 + 3 + 2 + 1 + 1 = 32 customers.
  3. Proportion: 32150 = 0.213, which is 21.3%.
  4. One in five is 20%. 21.3% is at least 20%, so the data supports the claim.
  5. It is only a little over 20%, and this is one sample of 150 customers, so a different sample could fall just below.
Full working
  1. 12 + 8 + 5 + 3 + 2 + 1 + 1 = 32rows from 30 minutes up
  2. 32150 = 0.213
  3. 0.213 ≥ 0.2
  4. 21.3% waited more than 30 minutes, so the data supports the claim
b
Merit

A triangular distribution with a = 5, c = 15 and b = 65 is suggested to model the waiting times.

(i) Justify the choice of each parameter. Include an observation from the table in your answer.

(ii) Use this model to estimate the probability that a customer waits more than 30 minutes.

Hint

The peak height of a triangle is 2b − a. The area to the right of 30 is a smaller triangle; find its height at 30 first.

Answer

(i) a = 5: the shortest wait was 5 minutes. b = 65: the longest wait was 64 minutes, so the waits end just below 65. c = 15: the two tallest classes are 10 to 15 (33) and 15 to 20 (30), so the most common wait is near 15 minutes. (ii) P(X > 30) = 0.4083.

Step by step
  1. (i) a is the minimum: no customer waited less than 5 minutes, so a = 5.
  2. b is the maximum: the longest wait was 64 minutes, inside the 60 to 65 class, so b = 65 covers every customer.
  3. c is the mode, the most likely wait. The tallest classes are 10 to 15 (33 customers) and 15 to 20 (30), so the peak is near 15 minutes.
  4. (ii) Let X = a customer's waiting time in minutes, X ~ triangular with a = 5, c = 15, b = 65.
  5. Peak height = 2b − a = 260 = 130.
  6. More than 30 minutes is the right tail: a triangle from 30 to 65.
  7. Its height at 30 is a fraction of the peak: 65 − 3065 − 15 × 130 = 3550 × 130 = 0.02333.
  8. Area = 12 × base × height = 12 × 35 × 0.02333 = 0.4083.
  9. Check with the formula: (b − x)²(b − a)(b − c) = 35²60 × 50 = 12253000 = 0.4083.
Full working
  1. a = 5 (shortest wait), c = 15 (tallest classes), b = 65 (longest wait 64)
  2. peak = 265 − 5 = 130
  3. f(30) = 65 − 3065 − 15 × 130 = 0.02333
  4. P(X > 30) = 12 × 35 × 0.02333right-tail triangle
  5. P(X > 30) = 0.4083
c
Merit

A staff member suggests changing the model to a = 5, c = 15 and b = 50, because only 4 customers waited 50 minutes or more.

Use this model to estimate the probability that a customer waits more than 30 minutes. Compare both triangular models with the data, and state which model fits this data better. Give one drawback of the model you choose.

Hint

Repeat part (b) with b = 50. Then put 0.4083, your new value and the data's 0.213 side by side.

Answer

With b = 50, P(X > 30) = 0.2540. The data gives 0.213. The b = 65 model gives 0.4083, almost double the data, so it badly overestimates long waits. The b = 50 model is much closer, so it fits better. Drawback: it gives zero probability of waiting 50 minutes or more, but 4 of 150 customers (2.7%) did.

Step by step
  1. New model: a = 5, c = 15, b = 50. Peak height = 250 − 5 = 245.
  2. Height at 30: 50 − 3050 − 15 × 245 = 2035 × 245 = 0.02540.
  3. P(X > 30) = 12 × 20 × 0.02540 = 0.2540.
  4. Now compare with the data. The data says 32150 = 0.213 of customers waited more than 30 minutes.
  5. The b = 65 model says 0.4083. That is nearly twice the data, so it overestimates long waits by a lot.
  6. The reason: the data falls away quickly after 30 minutes, but a triangle with b = 65 keeps a lot of area out to the right.
  7. The b = 50 model says 0.2540. It still overestimates, but only by about 4 percentage points, so it fits better.
  8. Drawback: the b = 50 model says no one can wait 50 minutes or more, but 4 customers did. It ignores the rare very long waits.
Full working
  1. peak = 250 − 5 = 245
  2. f(30) = 2035 × 245 = 0.02540
  3. P(X > 30) = 12 × 20 × 0.02540 = 0.2540
  4. data 0.213; b = 65 model 0.4083; b = 50 model 0.2540compare
  5. b = 50 fits better, but gives P(X ≥ 50) = 0 when 4 customers waited that long
d
Excellence

The station plans a promotion: if a customer waits longer than t minutes, their WOF is free. The manager wants only 5% of customers to get a free WOF.

(i) Use the triangular model with a = 5, c = 15 and b = 50 to find t, to 1 decimal place.

(ii) Use the table to estimate the proportion of customers who would actually wait longer than this t. Explain why it differs from 5%.

Hint

In the right tail, P(X > t) = (b − t)²(b − a)(b − c). Set that equal to 0.05 and solve for t.

Answer

(i) t = 41.1 minutes. (ii) About 11 of 150 customers, roughly 7%, waited longer than 41.1 minutes. The model has no waits past 50, but the data has a long right tail, so the model underestimates the proportion.

Step by step
  1. (i) t will be in the right tail, past the peak at 15. The right-tail area formula is (b − t)²(b − a)(b − c).
  2. (50 − t)²45 × 35 = 0.05, so (50 − t)² = 0.05 × 1575 = 78.75.
  3. 50 − t = √78.75 = 8.874, so t = 41.1 minutes.
  4. (ii) From the table, 3 + 2 + 1 + 1 = 7 customers waited 45 minutes or more.
  5. In the 40 to 45 class there are 5 customers. 41.1 to 45 is about 3.9 of those 5 minutes, so assume about 5 × 3.95 ≈ 3.9 of them waited more than 41.1 minutes.
  6. That gives about 7 + 3.9 ≈ 10.9 customers, and 10.9150 ≈ 0.072, roughly 7%.
  7. So the real proportion is higher than 5%. The b = 50 model cuts off the long tail: it says no one waits past 50, but 4 customers did.
  8. The model also fits the 30 minute tail a little high (part (c)), but the cut-off at 50 matters more this far out, so the free WOFs are underestimated.
Full working
  1. (50 − t)²45 × 35 = 0.05right-tail area
  2. (50 − t)² = 78.75, so t = 50 − 8.874 = 41.1
  3. data: 7 + 5 × 3.95 ≈ 10.9 customersspread evenly within 40 to 45
  4. 10.9150 ≈ 0.072
  5. t = 41.1 minutes; about 7% of customers in the data would get a free WOF, more than 5%, because the model leaves out waits over 50
e
Excellence

The station sees 200 customers a week and budgets for at most 15 free WOFs a week, using t from part (d).

Assuming 5% of customers get a free WOF, find the probability that the budget is exceeded in a week. Use your answer to part (d)(ii) to discuss whether the budget is safe, and state any assumption you make.

Hint

Each of the 200 customers gets a free WOF or not, so the number of free WOFs is binomial. Exceeded means more than 15.

Answer

X ~ Bin(200, 0.05): P(X > 15) = 0.0444. But the data suggests the rate is about 7%, and X ~ Bin(200, 0.07) gives P(X > 15) = 0.327, so the budget would be exceeded in about one week in three. It is not safe; raise t to about 45 minutes or budget for more. This assumes customers' waits are independent, which is doubtful: a busy morning makes many customers wait together, which makes bad weeks even more likely.

Step by step
  1. Let X = the number of free WOFs in a week of 200 customers. Each customer either gets one or not, with p = 0.05, assumed independent, so X ~ Bin(200, 0.05).
  2. The budget is exceeded when X > 15, which is P(X ≥ 16) = 1 − P(X ≤ 15) = 0.0444.
  3. On the model, the budget would be exceeded in only about 1 week in 22.
  4. But the data suggests the rate is nearer 7%. With X ~ Bin(200, 0.07), P(X > 15) = 0.327.
  5. So the budget would be exceeded in about one week in three. The budget is not safe; the manager should either raise t (about 45 minutes gives 7150 ≈ 4.7% in the data) or budget for more free WOFs.
  6. Also, waits on the same day are not independent: one busy morning makes many customers wait, so free WOFs come in clusters, which makes bad weeks even more likely.
Full working
  1. X ~ Bin(200, 0.05)X = free WOFs in a week, customers independent
  2. P(X > 15) = 1 − P(X ≤ 15) = 0.0444
  3. data rate ≈ 0.07 from part (d)(ii)
  4. X ~ Bin(200, 0.07): P(X > 15) = 0.327
  5. P(budget exceeded) = 0.0444 on the model, but about 0.33 at the data's rate, so the budget is not safe
Question Three

Question Three

a
Achieved

At a kiwifruit packhouse in Te Puke, 4% of kiwifruit are rejected for skin blemishes. A tray holds 30 kiwifruit.

Use a probability distribution model to estimate the probability that a tray has at most 2 rejected kiwifruit. State the name and parameters of the distribution as part of your answer.

Hint

A fixed number of fruit (30), each rejected or not, with the same chance (0.04): binomial.

Answer

X = number of rejected kiwifruit in a tray, X ~ Bin(30, 0.04). P(X ≤ 2) = 0.8831.

Step by step
  1. Define X = the number of rejected kiwifruit in a tray of 30.
  2. There is a fixed number of trials (30 fruit), two outcomes (rejected or not) and the same probability for each fruit (0.04).
  3. So X is binomial: X ~ Bin(30, 0.04), assuming fruit are rejected independently.
  4. At most 2 means 0, 1 or 2: P(X ≤ 2) = P(X = 0) + P(X = 1) + P(X = 2).
  5. These are 0.2939 + 0.3673 + 0.2219 = 0.8831.
  6. Calculator, binomial cdf with n = 30, p = 0.04, up to x = 2: 0.8831.
  7. Check: the mean is 30 × 0.04 = 1.2 rejects, so at most 2 should be very likely. 0.88 fits.
Full working
  1. X ~ Bin(30, 0.04)X = rejected fruit in a tray of 30
  2. P(X ≤ 2) = P(X = 0) + P(X = 1) + P(X = 2)
  3. = 0.2939 + 0.3673 + 0.2219
  4. P(X ≤ 2) = 0.8831
b
Merit

The random variable X is the number of trucks arriving at the packhouse loading dock in a 15-minute period during the harvest. Its probability distribution is shown in the table.

(i) Find the value of k.

(ii) Find the mean and standard deviation of X, and explain what the mean tells you in context.

(iii) The dock can unload 3 trucks in 15 minutes. Find the probability that at least one truck has to wait in a 15-minute period.

x012345
P(X = x)0.080.220.30k0.120.06
Hint

The probabilities add to 1. Then E(X) = Σ x × P(X = x), and Var(X) = Σ x² × P(X = x) − [E(X)]².

Answer

(i) k = 0.22. (ii) E(X) = 2.26 trucks, SD(X) = 1.309 trucks. Over many 15-minute periods, an average of 2.26 trucks arrive. (iii) P(X ≥ 4) = 0.18.

Step by step
  1. (i) All the probabilities must add to 1: 0.08 + 0.22 + 0.30 + k + 0.12 + 0.06 = 1.
  2. 0.78 + k = 1, so k = 0.22.
  3. (ii) Mean: multiply each value by its probability and add.
  4. E(X) = 0 × 0.08 + 1 × 0.22 + 2 × 0.30 + 3 × 0.22 + 4 × 0.12 + 5 × 0.06 = 0 + 0.22 + 0.60 + 0.66 + 0.48 + 0.30 = 2.26.
  5. For the variance, first find E(X²): 0 + 0.22 + 4 × 0.30 + 9 × 0.22 + 16 × 0.12 + 25 × 0.06 = 0.22 + 1.20 + 1.98 + 1.92 + 1.50 = 6.82.
  6. Var(X) = 6.82 − 2.26² = 6.82 − 5.1076 = 1.7124. SD(X) = √1.7124 = 1.309.
  7. In context: you can't have 2.26 trucks in one period, but over many 15-minute periods the average is 2.26 trucks.
  8. (iii) A truck has to wait if more than 3 arrive, which is 4 or 5 trucks.
  9. P(X ≥ 4) = 0.12 + 0.06 = 0.18.
Full working
  1. 0.78 + k = 1, so k = 0.22probabilities add to 1
  2. E(X) = 0.22 + 0.60 + 0.66 + 0.48 + 0.30 = 2.26
  3. E(X²) = 0.22 + 1.20 + 1.98 + 1.92 + 1.50 = 6.82
  4. Var(X) = 6.82 − 2.26² = 1.7124, SD = 1.309
  5. P(X ≥ 4) = 0.12 + 0.06 = 0.18
  6. k = 0.22; mean 2.26 trucks, SD 1.309 trucks; P(a truck waits) = 0.18
c
Merit

The packhouse's forklifts break down at an average rate of 0.6 breakdowns per week during the harvest.

(i) Use a Poisson distribution to find the probability of at least 2 breakdowns in a 4-week period.

(ii) Describe two conditions needed for the Poisson model in this context, and discuss whether one of them is likely to hold.

Hint

Scale the rate to 4 weeks first. At least 2 is 1 − P(X = 0) − P(X = 1).

Answer

(i) X ~ Po(2.4); P(X ≥ 2) = 0.6916. (ii) Breakdowns happen independently of each other, and at a constant average rate. A constant rate is doubtful: forklifts work harder in peak weeks and wear out, so breakdowns are likely to be more frequent at the height of the harvest.

Step by step
  1. (i) Define X = the number of forklift breakdowns in a 4-week period.
  2. The rate is 0.6 per week, so over 4 weeks λ = 0.6 × 4 = 2.4. So X ~ Po(2.4).
  3. At least 2 is everything except 0 and 1: P(X ≥ 2) = 1 − P(X = 0) − P(X = 1).
  4. P(X = 0) = e^(−2.4) = 0.0907. P(X = 1) = 2.4 × e^(−2.4) = 0.2177.
  5. P(X ≥ 2) = 1 − 0.0907 − 0.2177 = 0.6916.
  6. (ii) Condition 1, independence: one breakdown does not make another more or less likely.
  7. Condition 2, constant rate: breakdowns happen at the same average rate (0.6 a week) throughout the 4 weeks.
  8. Other conditions: breakdowns happen one at a time, never two at exactly the same moment, and at random.
  9. Discussion: the constant rate is doubtful. At the peak of the harvest the forklifts run for longer shifts with heavier loads, so they are more likely to break down then than early in the season.
  10. Independence may also fail: if a mechanic does a rushed repair, the same forklift may break down again soon.
Full working
  1. λ = 0.6 × 4 = 2.4scale to 4 weeks
  2. X ~ Po(2.4)X = breakdowns in 4 weeks
  3. P(X ≥ 2) = 1 − 0.0907 − 0.2177
  4. P(X ≥ 2) = 0.6916
  5. Conditions: independent breakdowns, constant rate; constant rate doubtful in peak harvest
d
Excellence

To check the binomial model from part (a), a quality controller counted the rejected kiwifruit in 200 trays of 30. The results are in the table.

(i) Find the mean and standard deviation of the number of rejects per tray in this data.

(ii) Use the model X ~ Bin(30, 0.04) to find the expected number of the 200 trays with 0 rejects and with 4 or more rejects.

(iii) Use these results to evaluate whether the binomial model is suitable. Give a reason in context for any difference between the model and the data.

Rejects in a tray0123456
Number of trays785432181062
Hint

Compare the data's mean and variance with the binomial's np and np(1 − p). For a binomial model, the variance must be a little less than the mean.

Answer

(i) Mean 1.27 rejects, SD 1.42 (variance 2.03). (ii) Expected 58.8 trays with 0 rejects (observed 78) and 6.1 trays with 4 or more (observed 18). (iii) The mean fits (1.27 against np = 1.2, so about 4% rejects), but the data's variance (2.03) is much larger than the binomial variance np(1 − p) = 1.152. The data has too many trays with no rejects and too many with lots. Fruit in a tray usually comes from the same orchard bin, so if one has blemishes (hail, wind rub) others are likely to as well: rejects are not independent and p is not constant between trays. The binomial model is not suitable; it underestimates the trays with many rejects.

Step by step
  1. (i) Mean: total rejects ÷ number of trays. Total = 0 × 78 + 1 × 54 + 2 × 32 + 3 × 18 + 4 × 10 + 5 × 6 + 6 × 2 = 254.
  2. Mean = 254 ÷ 200 = 1.27 rejects per tray.
  3. Calculator, 1-variable statistics with the frequencies: standard deviation = 1.42, so variance ≈ 2.03.
  4. (ii) From Bin(30, 0.04): P(X = 0) = 0.2939, so expected trays = 200 × 0.2939 = 58.8.
  5. P(X ≥ 4) = 1 − P(X ≤ 3) = 0.0306, so expected trays = 200 × 0.0306 = 6.1.
  6. The data has 78 trays with 0 rejects, and 10 + 6 + 2 = 18 trays with 4 or more.
  7. (iii) The centres agree. The binomial mean is np = 30 × 0.04 = 1.2, close to the data's 1.27, which is a reject rate of 1.27 ÷ 30 = 4.2%.
  8. The spreads do not. The binomial variance is np(1 − p) = 1.2 × 0.96 = 1.152 (SD 1.07), but the data's variance is about 2.03, nearly double.
  9. That shows up at both ends: many more trays with no rejects (78 against 58.8) and about three times as many trays with 4 or more (18 against 6.1).
  10. Reason in context: fruit in one tray is packed from the same orchard bin, and a bad block (hail damage, wind rub) gives many blemished fruit together. A clean orchard gives trays with none.
  11. So the binomial conditions fail: fruit in a tray are not independent, and p is not constant: it changes from orchard to orchard.
  12. Conclusion: the binomial model is not suitable. It gets the average right but underestimates the trays with many rejects, so if the packhouse re-sorts trays with 4 or more rejects, the model would badly underestimate that work.
Full working
  1. mean = 254200 = 1.27, SD = 1.42, variance ≈ 2.03data
  2. 200 × P(X = 0) = 200 × 0.2939 = 58.8observed 78
  3. 200 × P(X ≥ 4) = 200 × 0.0306 = 6.1observed 18
  4. np = 1.2, np(1 − p) = 1.152binomial mean and variance
  5. data variance 2.03 ≫ 1.152: too many trays with 0 and with 4+
  6. Binomial not suitable: rejects cluster by orchard, so independence and constant p fail
e
Excellence

A tray with 3 or more rejected kiwifruit is downgraded. The packhouse wants at least 95% of trays to have at most 2 rejects, and is considering a new grading machine that lowers the reject rate.

(i) Using a binomial model for trays of 30, find the largest reject rate p that meets the target, to 4 decimal places.

(ii) Using your findings from part (d), discuss whether a machine that achieves this reject rate would actually meet the target.

Hint

P(X ≤ 2) is a function of p. Set P(X ≤ 2) = 0.95 and solve for p using your calculator's solver or a table of values. P(X ≤ 2) gets smaller as p gets bigger.

Answer

(i) P(X ≤ 2) = (1 − p)³⁰ + 30p(1 − p)²⁹ + 435p²(1 − p)²⁸ = 0.95 gives p = 0.0278, so the reject rate must be at most 2.78% (down from 4%). (ii) Probably not. Part (d) showed rejects cluster by orchard, so trays vary more than the binomial model says. In the data, 78 + 54 + 32200 = 0.82 of trays had at most 2 rejects, below the binomial's 0.883 at a similar rate. So a 2.78% machine would likely give fewer than 95% of trays with at most 2 rejects; the rate would need to be lower still, or fruit from bad blocks mixed across trays.

Step by step
  1. (i) Let X = the number of rejects in a tray of 30, X ~ Bin(30, p). We need P(X ≤ 2) ≥ 0.95.
  2. Write it out: P(X ≤ 2) = (1 − p)³⁰ + 30p(1 − p)²⁹ + 435p²(1 − p)²⁸, where 435 = 30C2.
  3. At p = 0.04 this is 0.8831, too low. A larger p makes rejects more common, so P(X ≤ 2) falls as p rises. We need a smaller p.
  4. Solve P(X ≤ 2) = 0.95 on the calculator (or try values): p = 0.0278 gives 0.9501 and p = 0.0279 gives just under 0.95.
  5. So the largest reject rate is p = 0.0278, about 2.78%. The machine must cut the reject rate by about 30%.
  6. (ii) That answer relies on the binomial model: fruit rejected independently, with the same p for every tray.
  7. Part (d) showed this is not true. Rejects cluster, because a tray comes from one orchard bin, so the data's variance (2.03) was almost double the binomial's (1.152).
  8. More spread means more trays with many rejects. In the data, 78 + 54 + 32200 = 0.82 of trays had at most 2 rejects, while the binomial model at 4% says 0.883.
  9. So even if the machine averages 2.78%, clustering would push the share of trays with at most 2 rejects below 95%.
  10. To meet the target, the packhouse would need a lower reject rate than 2.78%, or to reduce the clustering, for example by mixing fruit from different bins in each tray or grading hail-damaged blocks separately.
Full working
  1. X ~ Bin(30, p), need P(X ≤ 2) ≥ 0.95
  2. (1 − p)³⁰ + 30p(1 − p)²⁹ + 435p²(1 − p)²⁸ = 0.95solve
  3. p = 0.0278: 0.9501; p = 0.0279: below 0.95
  4. data: P(X ≤ 2) = 164200 = 0.82 against 0.883 from the modelclustering
  5. p ≤ 0.0278; clustering means this rate would likely still give fewer than 95% of trays with at most 2 rejects

AS 91586 past papers · All NCEA Level 3 Maths

A Vertex Academy practice paper, written from NZQA's 2026 assessment specification for AS 91586. It is not an NZQA paper. Grades use the 2025 exam's cut scores: 8, 14 and 19 out of 24 for Achieved, Merit and Excellence.