This booklet covers every formula the 41 Cambridge 9709 Paper 6 (Probability and statistics 2) past papers in our bank needed. Paper 6 tests the Poisson distribution, linear combinations of random variables, continuous random variables, sampling and estimation, and hypothesis tests. Each row is marked Given or Learn it and lists the past questions that used it.
| Board | Official document |
|---|---|
| Cambridge International AS & A Level Mathematics 9709 | Cambridge 9709 Maths Paper 6 papers March 2020 to November 2025; List of formulae and statistical tables (MF19), printed in the 9709 syllabus for 2026 and 2027 |
Given means it is in the List of formulae and statistical tables (MF19) given in every paper. Learn it means a past question needed it and the exam did not print it.
The Poisson distribution
MF19 prints the Poisson probability, mean and variance. The rate scaling, sum and approximations are Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Poisson Po(λ): probabilityUse it for the probability of exactly r events; P(X = 0) is e-λ. For 'at least' or 'fewer than', add the terms or use 1 minus the complement, with no end errors. Setting two probabilities equal cancels the powers of λ. Used in: Nov 2025 paper 61 Q1(a), probability for Po(3); Jun 2025 paper 62 Q3(b), P(X = n) = P(X = n + 1); Mar 2020 paper 62 Q4(a), accidents in a year.Given | Given | |
| Poisson Po(λ): mean and varianceThe variance of a Poisson variable is its mean, so a Poisson with mean 200 has variance 200 and the normal approximation is N(200, 200). Used in: Nov 2025 paper 62 Q2(a), standard deviation of X + Y with Y ~ Po(4); Jun 2025 paper 62 Q6(b); Nov 2025 paper 61 Q1(c).Given | Given | |
| Mean number for a different periodScale the mean to the period in the question before using the Poisson formula: 14.4 per hour becomes 2.4 in 10 minutes. Used in: Nov 2025 paper 65 Q3(a), hits in a 10-minute period; Nov 2021 paper 62 Q5(a), 0.2 errors per page for 10 pages; Mar 2020 paper 62 Q4(a).Learn it | Learn it | |
| Sum of independent Poisson variablesAdd the means. The total of 100 values of X and 150 values of Y has mean 100 x 3 + 150 x 2 = 600 and is then approximated by N(600, 600). Used in: Nov 2025 paper 61 Q1(b), P(X + Y) for Po(3) and Po(2); Nov 2021 paper 62 Q5(c), typing and formatting errors; Jun 2025 paper 62 Q6(a).Learn it | Learn it | |
| Poisson approximation to a binomialUse it when n is large and p is small, then work with λ = np. Used in: Nov 2025 paper 61 Q3(a), 14 500 characters with p = 0.0001; Jun 2025 paper 62 Q6(a), B(700, 0.005); Nov 2025 paper 65 Q4(a), 2430 characters.Learn it | Learn it | |
| Conditions for the Poisson approximationState both conditions with the numbers; the Nov 2025 paper 65 Q4(a) scheme calls 'n is large and p is small' insufficient, and other schemes also accept p < 0.1 as the second condition. Used in: Nov 2025 paper 65 Q4(a), justify the approximating distribution; Nov 2025 paper 62 Q4(ab), why Poisson is not suitable for n = 40, p = 0.18; Jun 2022 paper 62 Q3(b).Learn it | Learn it | |
| Normal approximation to a PoissonMean and variance are both λ, so the standard deviation is the square root of λ. The specification gives the condition as λ large, about 15 or more. Used in: Nov 2025 paper 65 Q4(b), N(123.4, 123.4) for copies sold; Nov 2025 paper 62 Q1(c), 20-week period; Nov 2021 paper 62 Q5(b).Learn it | Learn it | |
| Continuity correctionX is the Poisson (or binomial) and Y the normal. Convert the words to whole numbers first: 'more than 50' is at least 51, so use 50.5. The standardising method mark does not need the correction, but the final accuracy mark does. Used in: Nov 2025 paper 65 Q4(b), 130.5 for more than 130; Nov 2025 paper 62 Q1(c), 50.5; Nov 2021 paper 62 Q5(b).Learn it | Learn it |
Linear combinations of random variables
MF19 prints none of these, so every row in this section is Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Linear function of one variableThe constant b changes the mean but not the variance, and the multiplier a is squared in the variance. Used in: Nov 2021 paper 62 Q1(b), price P = 11M + 0.50; Jun 2025 paper 61 Q4(a), cost at $0.40 per minute; Nov 2025 paper 65 Q5(b), price per kg.Learn it | Learn it | |
| Mean of a combinationIt holds whether or not X and Y are independent. For a difference such as L - 3S, use a = 1 and b = -3. Used in: Nov 2025 paper 61 Q6(b), large bag against 3 small bags; Mar 2020 paper 62 Q6(b), L - 2S; Nov 2025 paper 62 Q2(b).Learn it | Learn it | |
| Variance of a combinationThe variances add even for a difference, because a minus sign is squared: Var(X - Y) = Var(X) + Var(Y). Used in: Nov 2025 paper 62 Q2(b), standard deviation of 5X - Y; Jun 2025 paper 62 Q1(b), difference between two students' times; Mar 2020 paper 62 Q6(b).Learn it | Learn it | |
| Sum of n independent observationsThis is different from n times one value, where the variance is n2 σ2. Say that the observations are independent. Used in: Jun 2025 paper 61 Q4(b), total paid by 35 people; Jun 2025 paper 63 Q5, n men on a footbridge.Learn it | Learn it | |
| Combination of independent normal variablesIn N(mean, variance) the second number is the variance, so N(2.5, 0.05) has standard deviation sqrt(0.05). Work out the mean and variance of the combination, then standardise. Used in: Nov 2025 paper 61 Q6(a), total mass of a large and a small bag; Nov 2025 paper 65 Q5(a); Mar 2020 paper 62 Q6(a).Learn it | Learn it | |
| StandardisingUse the standard deviation, the square root of the variance. For P(difference > 12) with a difference that can be either way, double the one-sided probability. Used in: Jun 2025 paper 62 Q1(b), probability the difference exceeds 12 minutes; Nov 2025 paper 61 Q6(b); Jun 2025 paper 63 Q5.Learn it | Learn it |
Continuous random variables
MF19 prints E(X), Var(X) and the integration rules. The total area, probabilities and the median from a density function are Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Total area under a density functionUse it to find an unknown constant such as k, then keep that value for later parts. Used in: Jun 2025 paper 62 Q7(a), show that k = 3/a3; Nov 2025 paper 62 Q5(a); Nov 2021 paper 62 Q7(a).Learn it | Learn it | |
| Probability from a density functionIntegrate between the limits; an upper tail P(X > c) runs from c to the top of the domain. Used in: Nov 2025 paper 61 Q7(a), time longer than 4.5 minutes; Jun 2025 paper 61 Q7(a); Mar 2020 paper 62 Q5(a), bottle containing more than 305 ml.Learn it | Learn it | |
| Median of a continuous variablem is the median; start the integral at the lower end of the domain. For a symmetrical density function the median is at the line of symmetry. Used in: Jun 2025 paper 62 Q7(c), median of X; Nov 2023 paper 62 Q5(c); Nov 2025 paper 62 Q5(b), P(X <= m) = 0.5.Learn it | Learn it | |
| Expectation of a continuous random variableIntegrate x f(x) over the domain. Use it to find an unknown constant when E(X) is given, as well as to find E(X). Used in: Jun 2025 paper 62 Q7(b), find a given E(X); Nov 2025 paper 65 Q6(bc), find a from E(X); Nov 2024 paper 61 Q4(b), show that E(X) involves a log.Given | Given | |
| Variance of a continuous random variableIntegrate x2 f(x), then subtract the square of E(X). Used in: Mar 2024 paper 62 Q6(c), find Var(X).Given | Given | |
| Power of xIntegrate each power of x: add 1 to the power and divide by the new power. A density function with brackets is expanded first. Used in: Nov 2025 paper 61 Q7(a); Nov 2025 paper 62 Q5(a); Jun 2025 paper 62 Q7(a).Given | Given | |
| One over xA term in 1/x integrates to ln of x, which appears when x f(x) contains a 1/x term. Used in: Nov 2024 paper 61 Q4(b), show that E(X) involves ln; Nov 2023 paper 62 Q5(a), show that b = 2a.Given | Given | |
| cos xIntegrating a density function that contains cos x gives sin x. Used in: Jun 2025 paper 61 Q7(a), show a probability for a density function with cos x.Given | Given | |
| Integration by partsUse it for E(X) when f(x) contains cos x, because x cos x needs integration by parts. Used in: Jun 2025 paper 61 Q7(b), show that E(X) takes a given value.Given | Given |
Sampling and estimation
MF19 prints the unbiased estimates and the distribution of a sample mean and of a sample proportion. Confidence intervals are Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Unbiased estimate of the meanThe unbiased estimate of the population mean is the sample mean. Used in: Nov 2021 paper 62 Q1(a), estimates from 40 blocks of cheese; Mar 2020 paper 62 Q3(a); Jun 2025 paper 61 Q3(a).Given | Given | |
| Unbiased estimate of the varianceDivide by n - 1, not n: the schemes give no marks for the biased variance. The second form needs only the totals of x and x2, and rearranging it gives the sum of x2 from a known estimate. Used in: Nov 2021 paper 62 Q1(a); Jun 2025 paper 62 Q2, find the sum of h2; Mar 2020 paper 62 Q3(a).Given | Given | |
| Central Limit TheoremThe mean of a sample of n values has mean μ and variance σ2/n, so standardise with σ divided by the square root of n. It is normal if the population is normal, or approximately normal for a large sample. Used in: Nov 2025 paper 61 Q2, test on 60 packets of tea; Jun 2025 paper 61 Q2(ab), sample of 100 values from a binomial; Jun 2025 paper 63 Q2(b).Given | Given | |
| Confidence interval for a meanUse the population standard deviation if it is known; if not, use the square root of the unbiased variance for a large sample. The answer must be an interval. Used in: Nov 2025 paper 65 Q1(ab), 96% interval from 80 lectures; Jun 2025 paper 65 Q2(ab); Nov 2025 paper 62 Q3(a).Learn it | Learn it | |
| Width of a confidence intervalSet the width equal to the value given and solve for z, σ or n. A higher confidence level or a smaller sample gives a wider interval. Used in: Mar 2020 paper 62 Q2, find the confidence level from a width of 1.25 cm; Nov 2025 paper 61 Q4(a), width ratio 1.414; Jun 2025 paper 63 Q3(ab), wider or narrower.Learn it | Learn it | |
| Confidence interval for a proportionp-hat is the sample proportion, the number found divided by n; this uses the approximate distribution of a sample proportion printed in MF19 (S20.3), with p-hat in place of p. Used in: Jun 2025 paper 61 Q6(c), 5 of 40 students; Nov 2021 paper 62 Q3(a), 35 reds in 140 spins; Jun 2025 paper 62 Q4, width 0.1066.Learn it | Learn it | |
| z value for a confidence levelFor a c% interval, look up the z that leaves (100 - c)/2 % in each tail: 90% gives 1.645, 95% gives 1.960 and 99% gives 2.576. The schemes accept z = 2.054 or 2.055 for 96%. Used in: Nov 2025 paper 65 Q1(ab); Nov 2021 paper 62 Q3(a); Mar 2020 paper 62 Q2, confidence level from z = 1.953.Learn it | Learn it | |
| All of several confidence intervals contain the valueIndependent c% intervals each contain the true value with probability c/100; for exactly some of them, use a binomial or add the cases. Used in: Nov 2025 paper 62 Q3(c), largest r with probability over 0.5; Nov 2021 paper 62 Q3(b), exactly two of three intervals.Learn it | Learn it |
Hypothesis tests
MF19 prints the binomial probability and the distribution of a sample mean. The test statistic and the error probabilities are Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Test statistic for a meanThis is the standardised form of the sample mean distribution in MF19 (S20.2), with mu0 from H0. Use the population standard deviation, or the square root of the unbiased variance for a large sample, and divide by the square root of n. Used in: Nov 2025 paper 62 Q6(a), mean weekly profit; Nov 2025 paper 61 Q2; Nov 2021 paper 62 Q4(bc), fireworks at the 1% level.Learn it | Learn it | |
| Critical value of the sample meanRearrange the test statistic to find the boundary of the rejection region, using the z from the table for the significance level. Used in: Nov 2025 paper 62 Q6(c), critical value a before the Type II error; Jun 2025 paper 62 Q5, smallest possible sample mean.Learn it | Learn it | |
| Binomial B(n, p): probabilityFor a test on a proportion with a small sample, X ~ B(n, p0) under H0. Add the tail probabilities and compare the tail probability with the significance level, with no end errors. Used in: Jun 2025 paper 62 Q8(a), 2 sixes in 30 throws; Nov 2025 paper 61 Q5(a), 40 households; Nov 2021 paper 62 Q6(b), rejection region.Given | Given | |
| Probability of a Type I errorUse the null value of the parameter and add the probabilities of every value in the rejection region. For a region found at the stated significance level it is no more than that level. Used in: Nov 2025 paper 61 Q5(a), find the probability of a Type I error; Jun 2025 paper 62 Q8(b); Nov 2021 paper 62 Q6(c).Learn it | Learn it | |
| Probability of a Type II error (small sample)Use the true value of the parameter given, not the null value, and add the probabilities of the values outside the rejection region. Used in: Jun 2025 paper 62 Q8(c), true p = 0.02; Nov 2021 paper 62 Q6(d), a 0 once in 40 digits; Mar 2020 paper 62 Q7(c).Learn it | Learn it | |
| Probability of a Type II error (test on a mean)Find the critical value of the sample mean first, then standardise it with the true mean and take the tail on the acceptance side. Used in: Nov 2025 paper 62 Q6(c), true mean $718.Learn it | Learn it |
Learn these: not in the booklet
- Conditions for a Poisson modelEvents occur singly, at random and independently of each other, at a constant average rate; give one in the context of the question. Used in: Nov 2025 paper 65 Q4(b); Nov 2025 paper 62 Q1(a).
- At least or fewer than for a PoissonUse P(X >= k) = 1 - P(X <= k - 1), and write out every term of the Poisson probability with no end errors. Used in: Nov 2025 paper 65 Q3(a); Nov 2025 paper 65 Q4(a); Mar 2020 paper 62 Q1.
- No events in a periodP(X = 0) = e-λ, so a condition such as 'no accidents in n days with probability more than 0.95' becomes e-λ > 0.95; take natural logs and solve for n. Used in: Mar 2020 paper 62 Q4(b).
- Straight-line density functionThe area of a triangle, one half x base x height, equals 1; this gives a relationship between the constants. Used in: Nov 2025 paper 65 Q6(a).
- Symmetrical density functionThe median is at the line of symmetry. Used in: Nov 2025 paper 61 Q7(bc); Mar 2020 paper 62 Q5(c).
- Random samplingEvery member of the population must have a chance of selection; number the members and use random numbers, and a method that reaches only some of the population is unsatisfactory. Used in: Nov 2021 paper 62 Q2(a); Nov 2021 paper 62 Q2(b); Jun 2025 paper 62 Q1(a).
- When the Central Limit Theorem is neededIt is needed when the population is not normal, as for a binomial, and the sample is large; it is not needed when the population is normal. Used in: Jun 2025 paper 61 Q2(ab); Jun 2025 paper 63 Q2(b); Nov 2025 paper 62 Q3(b).
- HypothesesWrite H0 and H1 in terms of the population parameter, such as the population mean or the proportion p, with p defined, not just 'mean'. Used in: Nov 2025 paper 65 Q2(b); Nov 2025 paper 61 Q2; Jun 2025 paper 61 Q6(a).
- One-tailed or two-tailedUse a two-tailed test when the claim is 'different from' and a one-tailed test for 'less than' or 'more than'. Used in: Nov 2025 paper 65 Q2(a); Nov 2025 paper 62 Q6(a).
- Conclusion of a testCompare the test statistic with the critical value (or the probability with the significance level), then say 'insufficient evidence that ...' or 'sufficient evidence that ...' in the context, never definite. Used in: Nov 2025 paper 65 Q2(c); Nov 2025 paper 62 Q6(a); Nov 2025 paper 61 Q2.
- A necessary assumptionFor a test on a mean with a known σ, assume the standard deviation is unchanged; for a total of independent values, assume independence. Used in: Nov 2025 paper 62 Q6(a); Jun 2025 paper 62 Q5; Jun 2025 paper 63 Q5.
- Rejection region and the errorsA Type I error is rejecting H0 when it is true and a Type II error is not rejecting it when it is false; if H0 is rejected a Type II error is impossible, and if it is not rejected a Type I error is impossible. Used in: Nov 2025 paper 61 Q5(c); Nov 2025 paper 65 Q3(b); Nov 2025 paper 61 Q5(b).
Download the booklet
This page as a printable PDF, with the full CuratedPrep study booklet for the course. Open the dropdown, then click a title to open it or use Download to save the PDF.
Cambridge1 booklet
How to use the formula sheet in the exam
- State both conditions for a Poisson approximation to a binomial, n > 50 and np < 5, with the numbers; the Nov 2025 paper 65 Q4(a) scheme calls 'n large and p small' insufficient.
- In N(mean, variance) the second number is the variance. Take its square root before standardising, and add variances, never standard deviations, for a combination.
- For an unbiased variance divide by n - 1. The Nov 2021 paper 62 Q1(a) scheme gives M0 A0 for the biased variance, and a biased value costs the accuracy marks in tests and intervals.
- In a test, write the hypotheses in terms of the population parameter, show the test statistic with σ over the square root of n, compare it with the critical value from the normal table, and conclude in context without being definite.
- For a Type II error use the true parameter value given in the question, and for a test on a mean find the critical value of the sample mean first.
- Know the critical values from the table of normal critical values: 1.645 for a one-tailed 5% test, 1.960 for a two-tailed 5% test and 2.326 for a one-tailed 1% test; a 96% interval needs z = 2.054 from the normal distribution function table.
Common questions
Which formulae does MF19 give me in Cambridge 9709 Paper 6?
Paper 6 comes with the whole MF19 booklet. Its sheet prints the Poisson probability with its mean and variance, E(X) and Var(X) for a continuous variable, the unbiased estimates of the mean and variance, the distribution of a sample mean (the Central Limit Theorem), the approximate distribution of a sample proportion, the binomial formulas, the integration rules, and the normal tables. It does not print the confidence interval formulas, the test statistic or the Type I and Type II error probabilities.
Do I need the cumulative distribution function in Paper 6?
No. The specification says explicit knowledge of the cumulative distribution function is not included. Find probabilities and the median by integrating the density function directly.
Do I have to prove the Poisson or the linear combination results?
No. The specification says proofs are not required for the Poisson mean and variance or for the linear combination results; you need to use them.
When do I use a normal approximation to a Poisson?
When the mean λ is large, about 15 or more according to the specification. Use N(λ, λ) with a continuity correction.
Other papers
Paper 1 · Paper 2 · Paper 3 · Paper 4 · Paper 5. All Cambridge 9709 paper formula sheets.
Practise this topic
Real past paper questions, by topic
CuratedPrep worksheets group real past paper questions by syllabus topic, with a full marking scheme and revision notes, for every Cambridge A Level maths 9709 topic.
In Dubai or the UAE? Work one to one with a CuratedPrep maths tutor.
Formulas checked against the official documents linked above. Boards update their booklets from time to time: always practise with the version your exam uses.