Skip to content
Marlbridge

Study Guides

Cambridge International AS & A Level Mathematics 9709: Probability & Statistics 2 – Study Guide

Study guide for Cambridge 9709 Probability & Statistics 2 (Paper 6): Poisson, linear combinations, PDFs, estimation and hypothesis tests, fully worked.

Subject
Mathematics
Level
A LEVELS
Topic
Probability & Statistics 2
Updated

Aligned to Cambridge A Level Mathematics (9709), 2026-2027. Official specification .

Syllabus page (what it covers and how it is assessed): Cambridge A Level Mathematics.

Syllabus points this page covers

9709

  • 6.1 The Poisson distribution
  • 6.2 Linear combinations of random variables
  • 6.3 Continuous random variables
  • 6.4 Sampling and estimation
  • 6.5 Hypothesis tests

This page covers all 5 subtopics recorded here in topic 6, Probability & Statistics 2.

Found an error? Report a correction.

Need help with this topic? Request a free trial class for A Level Mathematics (9709).

This guide teaches topic 6, Probability & Statistics 2, of the Cambridge International AS & A Level Mathematics 9709 syllabus for exams in 2026 and 2027 (Version 4). It covers syllabus sections 6.1 to 6.5. This content is examined on Paper 6: 1 hour 15 minutes, 50 marks, 6 to 8 structured questions, worth 20% of the A Level. Paper 6 is offered only as part of the full A Level, and it cannot be combined with Paper 4 (Mechanics). The syllabus assumes you know all of Probability & Statistics 1 and the calculus in Pure Mathematics 3. A scientific calculator is allowed, and you get MF19 (formulae and statistical tables).

Related pages: the Probability & Statistics 2 revision notes, the Probability & Statistics 2 practice questions, the Probability & Statistics 1 study guide, the A Level Mathematics hub, the printable 9709 checklist and the free A Level diagnostic.

What this unit covers

Section What you must be able to do Paper
6.1 The Poisson distribution Po(λ) probabilities; mean = variance = λ; model random events; approximate B(n, p) by Poisson; approximate Po(λ) by a normal 6
6.2 Linear combinations of random variables E and Var of aX + b and aX + bY; normal and Poisson sums 6
6.3 Continuous random variables Properties of a probability density function (single interval); probabilities, mean, variance, median and percentiles 6
6.4 Sampling and estimation Samples and randomness; distribution of X̄; Central Limit Theorem; unbiased estimates; confidence intervals for a mean and a proportion 6
6.5 Hypothesis tests One- and two-tailed tests for binomial, Poisson and normal means; Type I and Type II errors 6

6.1 The Poisson distribution

X ~ Po(λ) models the number of events in a fixed interval of time or space when events occur singly, independently, at random, and at a constant average rate. From MF19:

P(X = r) = e^(−λ) λʳ / r!        mean = λ        variance = λ

If the rate is λ per unit, the count in t units is Po(λt). A mean and variance that are roughly equal in data is evidence that a Poisson model may fit.

Worked example. Calls reach a help desk at random at a mean rate of 3.2 per hour.

X ~ Po(3.2)
P(X = 2) = e^(−3.2) × 3.2² / 2 = 0.209 (3 s.f.)
P(X ≥ 3) = 1 − e^(−3.2)(1 + 3.2 + 3.2²/2) = 0.620 (3 s.f.)
P(no calls in 30 minutes): Y ~ Po(1.6), P(Y = 0) = e^(−1.6) = 0.202 (3 s.f.)

Poisson approximation to the binomial

If X ~ B(n, p) with n large and p small (the syllabus gives n > 50 and np < 5, approximately), use X ≈ Po(np).

Worked example. 1.5% of items are faulty; a box holds 200. n = 200 > 50 and np = 3 < 5, so use Po(3).

P(at most 2 faulty) ≈ e^(−3)(1 + 3 + 9/2) = 8.5e^(−3) = 0.423 (3 s.f.)
Paired bars for r = 0 to 9 compare the exact binomial B(200, 0.015) with Po(3); they are almost identical. P(X ≤ 2) is 0.421 exactly and 0.423 by the Poisson. The rule is n > 50 and np < 5.
B(200, 0.015) and Po(3) side by side, showing why Po(np) works for large n and small p.

Normal approximation to the Poisson

If λ is large (λ > 15, approximately), use X ≈ N(λ, λ) with a continuity correction.

Worked example. X ~ Po(25). Find P(X > 30).

N(25, 25):  P(X > 30) = P(X ≥ 31) ≈ P(Y > 30.5)
z = (30.5 − 25)/5 = 1.1
P = 1 − Φ(1.1) = 1 − 0.8643 = 0.136 (3 s.f.)
Bar chart of Po(3.2) for r = 0 to 10. The bar at r = 2 is gold with probability 0.209 and the bars from r = 3 upward are dark teal with total 0.620. A dashed line at 3.2 marks the mean, which equals the variance.
The Po(3.2) distribution, with P(X = 2) = 0.209 and P(X ≥ 3) = 0.620 marked.
Bars for Po(25) with the curve N(25, 25) drawn over them. Bars from r = 31 are gold and a dashed line at 30.5 shows the continuity-corrected boundary. z = 1.1 gives 1 − Φ(1.1) = 0.136.
Why P(X > 30) becomes P(Y > 30.5) when a Poisson is approximated by a normal.
Bell curve for N(250, 4²), symmetric about 250 g. Shaded bands show 68% within 1 sd (246 to 254), 95% within 2 sd and 99.7% within 3 sd. The points of inflection at 246 and 254 are marked, and a second row gives z from −3 to 3.
The Normal curve is symmetrical, has inflection points at μ ± σ and holds about 68%, 95% and 99.7% of values within 1, 2 and 3 sd.

6.2 Linear combinations of random variables

These results are not in MF19. Learn them; proofs are not required.

E(aX + b) = aE(X) + b              Var(aX + b) = a² Var(X)
E(aX + bY) = aE(X) + bE(Y)
Var(aX + bY) = a² Var(X) + b² Var(Y)     (X, Y independent)

Variances always add, even for a difference: Var(X − Y) = Var(X) + Var(Y).

  • If X is normal, aX + b is normal.
  • If X and Y are independent normals, aX + bY is normal.
  • If X and Y are independent Poissons, X + Y is Poisson with mean λ₁ + λ₂.

Two separate items or one item scaled? The total mass of 4 separate bags is X₁ + X₂ + X₃ + X₄, with variance 4σ². Four times the mass of one bag is 4X, with variance 16σ².

Worked example. Bags of flour have mass X ~ N(1010, 8²) grams. A crate has mass Y ~ N(150, 5²) grams. Four bags are packed into one crate. Find the probability that the total mass exceeds 4200 g.

T = X₁ + X₂ + X₃ + X₄ + Y
E(T) = 4 × 1010 + 150 = 4190
Var(T) = 4 × 64 + 25 = 281
P(T > 4200): z = (4200 − 4190)/√281 = 0.5965
P = 1 − Φ(0.5965) = 0.275 (3 s.f.)
Two normal curves share mean 4040 g. The tall narrow one is the total of four separate bags (variance 4σ² = 256) and the low wide one is four times one bag (variance 16σ² = 1024).
Adding four independent bags gives variance 4σ², but scaling one bag by 4 gives 16σ².

6.3 Continuous random variables

A probability density function f(x) defined on one interval must satisfy f(x) ≥ 0 and ∫ f(x) dx = 1 over that interval. Probabilities are areas: P(a < X < b) = ∫ from a to b of f(x) dx. For a continuous variable, P(X = a) = 0, so < and ≤ give the same answer. From MF19:

E(X) = ∫ x f(x) dx        Var(X) = ∫ x² f(x) dx − {E(X)}²

The median m solves ∫ from the lower limit to m of f(x) dx = 1/2. Other percentiles work the same way. The syllabus excludes explicit use of the cumulative distribution function, so find these by direct integration of f(x).

Worked example. f(x) = k(4x − x²) for 0 ≤ x ≤ 3, and 0 otherwise.

∫₀³ (4x − x²) dx = [2x² − x³/3]₀³ = 18 − 9 = 9,  so k = 1/9
E(X)  = (1/9) ∫₀³ (4x² − x³) dx = (1/9)(36 − 81/4) = 7/4
E(X²) = (1/9) ∫₀³ (4x³ − x⁴) dx = (1/9)(81 − 243/5) = 18/5
Var(X) = 18/5 − (7/4)² = 0.5375 = 0.538 (3 s.f.)
median: (1/9)(2m² − m³/3) = 1/2  →  m³ − 6m² + 13.5 = 0
        root in [0, 3]: m = 1.79 (3 s.f.)

Infinite domain. The syllabus allows domains such as x ≥ 2. If f(x) = k/x⁴ for x ≥ 2, then ∫₂^∞ kx⁻⁴ dx = k/24, so k = 24. E(X) = ∫₂^∞ 24x⁻³ dx = 3. The median solves 1 − 8/m³ = 1/2, so m³ = 16 and m = 2.52 (3 s.f.).

Graph of f(x) = (4x − x²)/9 on 0 to 3. The total area is 1 and a dashed line at the median 1.79 splits it into two halves. The mean, 7/4, sits just left of the median. Variance is 0.538.
A pdf has area 1; the median splits the area in half and the mean is the balance point.

6.4 Sampling and estimation

A population is the whole group; a sample is the part you observe. A sample should be random so that every member has an equal chance of selection and the results are not biased. A method is unsatisfactory if some members cannot be chosen (for example, only surveying people at one place or time). Random numbers can produce a random sample: number the population, generate random numbers, and select the matching members, ignoring repeats and numbers out of range. Named methods such as quota or stratified sampling are not required.

The sample mean

X̄ is a random variable with E(X̄) = μ and Var(X̄) = σ²/n.

  • If X is normal, X̄ is exactly normal: X̄ ~ N(μ, σ²/n).
  • If X is not normal but n is large, the Central Limit Theorem says X̄ is approximately N(μ, σ²/n).

Worked example. A population has mean 30 and standard deviation 6. A random sample of 50 is taken. n is large, so by the CLT X̄ ≈ N(30, 36/50).

P(X̄ > 31): z = (31 − 30)/(6/√50) = 1.179
P = 1 − Φ(1.179) = 0.119 (3 s.f.)

Unbiased estimates

From MF19:

x̄ = Σx / n        s² = (1/(n − 1)) ( Σx² − (Σx)²/n )

“Unbiased” means that although individual estimates vary, the method gives the right value on average.

Confidence intervals

For a mean with known σ (normal population), or a large sample (use s):

x̄ ± z × σ/√n

For a proportion from a large sample, with p̂ the sample proportion:

p̂ ± z √( p̂(1 − p̂)/n )

Common z values from the critical values table: 90% → 1.645, 95% → 1.960, 98% → 2.326, 99% → 2.576.

Worked example. For a sample of 40 values, Σx = 2260 and Σx² = 128 400. Find a 95% confidence interval for μ.

x̄ = 2260/40 = 56.5
s² = (128400 − 2260²/40)/39 = 710/39 = 18.21
interval: 56.5 ± 1.96 × √(18.21/40) = 56.5 ± 1.322
          (55.2, 57.8)

The sample is large, so the CLT means you do not need to assume the population is normal.

Proportion. In a sample of 200, 72 have a property. p̂ = 0.36. A 90% interval is 0.36 ± 1.645 × √(0.36 × 0.64/200) = 0.36 ± 0.0558, giving (0.304, 0.416).

Two normal curves centred on 30: a wide one for a single value (sd 6) and a narrow one for the mean of 50 values (sd 0.85). The area of the narrow curve above 31 is shaded and equals 0.119.
The mean of 50 values varies far less than one value, so P(X̄ > 31) is only 0.119.
A normal curve of the sample mean centred on 56.5 with standard error 0.675. The central 95% of area, from 55.2 to 57.8, is shaded teal with 2.5% in each tail, and a bracket marks the interval (55.2, 57.8).
The 95% interval is the sample mean plus or minus 1.96 standard errors.

6.5 Hypothesis tests

Five-step right-tailed Poisson significance test using the lesson data, with Type I and Type II errors.
State hypotheses, model, tail, decision and contextual conclusion. Select the image to enlarge it.

Vocabulary: null hypothesis H₀ (the parameter has its usual value), alternative hypothesis H₁ (it has changed: <, > or ≠), significance level, test statistic, rejection (critical) region, acceptance region. A one-tailed test looks for change in one direction; a two-tailed test splits the significance level between both tails. Conclusions must be in context and not overstated: “there is evidence at the 5% level that…” or “there is insufficient evidence that…”.

Binomial, direct evaluation. A claim says p = 0.3. In 20 trials there are 2 successes. Test H₀: p = 0.3 against H₁: p < 0.3 at the 5% level.

X ~ B(20, 0.3) under H₀
P(X ≤ 2) = 0.0355 < 0.05
Reject H₀: evidence at the 5% level that p is less than 0.3.

Poisson, direct evaluation. A shop sells a mean of 4.5 umbrellas per week. In one week after an advert it sells 9. Test at 5% whether the mean has increased.

H₀: λ = 4.5    H₁: λ > 4.5
P(X ≥ 9) = 1 − P(X ≤ 8) = 0.0403 < 0.05
Reject H₀: evidence at the 5% level that mean weekly sales have increased.

For large n or large λ, the syllabus also allows a normal approximation, with a continuity correction, in place of direct evaluation.

Normal mean. Lengths have σ = 12 cm. H₀: μ = 500, H₁: μ ≠ 500, 5% level, sample of 36 with x̄ = 496.

z = (496 − 500)/(12/√36) = −2.000
two-tailed critical values ±1.960; −2.000 < −1.960
Reject H₀: evidence at the 5% level that the mean length is not 500 cm.

Type I and Type II errors

  • Type I error: rejecting H₀ when it is true. P(Type I) = probability of landing in the rejection region when H₀ is true. For a discrete test, find the actual value, which is usually below the stated significance level.
  • Type II error: accepting H₀ when it is false. You need a specific alternative value of the parameter.

Worked example (Poisson). In the umbrella test, P(X ≥ 8) = 0.0866 and P(X ≥ 9) = 0.0403 under H₀, so the rejection region is X ≥ 9 and P(Type I) = 0.0403. If the true mean is now 7, P(Type II) = P(X ≤ 8 | λ = 7) = 0.729 (3 s.f.).

Worked example (normal). In the lengths test, H₀ is accepted when 500 − 1.96 × 2 < x̄ < 500 + 1.96 × 2, that is 496.08 < x̄ < 503.92. If in fact μ = 505:

P(Type II) = Φ((503.92 − 505)/2) − Φ((496.08 − 505)/2)
           = Φ(−0.54) − Φ(−4.46) = 0.295 (3 s.f.)
Two bar charts share a dashed boundary between 8 and 9. Under H0 (Po(4.5)) the bars from 9 upward have probability 0.0403, the Type I error. If the true mean is 7 (Po(7)) the bars up to 8 have probability 0.729, the Type II error.
Type I error is landing in the rejection region when H0 is true; Type II is landing outside it when H0 is false.

Common errors

  • Rate not scaled to the interval (a rate per hour used for a 30-minute count).
  • Subtracting variances for X − Y.
  • Using 4X for four independent items.
  • Forgetting to check that k makes the total area 1 before finding E(X).
  • Dividing by n instead of n − 1 for an unbiased variance estimate.
  • Using σ²/n for a proportion instead of p̂(1 − p̂)/n.
  • Comparing a single probability, P(X = 9), with 5% instead of the tail P(X ≥ 9).
  • A conclusion that “proves” H₁, or that has no context.

Frequently asked questions

Can you take Cambridge 9709 Paper 6 with Paper 4? No. Paper 6 is offered only as part of the full A Level, and it cannot be combined with Paper 4 (Mechanics). It assumes all of Probability & Statistics 1 and the calculus in Pure Mathematics 3.

What conditions does a Poisson model need? Events must occur singly, independently, at random, and at a constant average rate in a fixed interval of time or space. For a Poisson distribution the mean equals the variance.

How do you find the median of a continuous random variable? Solve the equation that sets the integral of f(x) from the lower limit up to m equal to 1/2. The 9709 syllabus excludes explicit use of the cumulative distribution function, so integrate f(x) directly.

What does the Central Limit Theorem say? For a large sample, the sample mean X̄ is approximately normally distributed with mean μ and variance σ²/n, even if the population is not normal. If the population is normal, X̄ is exactly normal for any n.

What is a Type II error? A Type II error is accepting H₀ when it is false. To find its probability you need a specific alternative value of the parameter.

Next steps

Test yourself with the Probability & Statistics 2 practice questions, then keep the revision notes for the final weeks.

Official syllabus

Cambridge International AS & A Level Mathematics 9709 syllabus, for exams in 2026 and 2027 (Version 4), Cambridge University Press & Assessment. Topic 6, Probability & Statistics 2 (for Paper 6): sections 6.1 to 6.5.

Get free revision emails (optional)

Occasional emails with practice questions, worked explanations and links to free resources for the qualification and subjects you choose. No spam, and you can unsubscribe from any email. The free tools on this site never need an email.

Subjects (optional, up to 6)

Choose a qualification to see its subjects.

Related resources

Related articles

Studying this with a teacher

Working through Mathematics A LEVELS?

This page is free and stays free. If you would rather be taught it, Marlbridge runs Mathematics classes one-to-one and in small groups of up to 15, online in your own time zone. The first trial class is free. WhatsApp replies within an hour (8am–11pm Pakistan time, every day); email the same day.