This booklet covers every formula the 9 Edexcel International A Level Mathematics Unit S3 papers in our bank needed, from June 2022 to January 2026. S3 tests combinations of random variables, sampling, estimators, confidence intervals, hypothesis tests on means, goodness of fit and contingency tables, and rank correlation. Each row is marked Given or Learn it and lists the past questions that used it.
| Board | Official document |
|---|---|
| Pearson Edexcel International A Level Mathematics | Edexcel IAL Maths Unit S3 (WST03) papers June 2022 to January 2026; Mathematical Formulae and Statistical Tables, Issue 2, January 2021 (Pearson Edexcel International Advanced Subsidiary/Advanced Level in Mathematics, Further Mathematics and Pure Mathematics) |
Given means it is in the Mathematical Formulae and Statistical Tables booklet (Issue 2, 2021) handed out in every unit exam. Learn it means a past question needed it and the exam did not print it.
Combinations of random variables
The booklet prints the variance rule for independent variables. The specification lists the Normal result for aX plus or minus bY as a formula you are expected to know, so it is Learn it.
| Formula | Equation | In the exam |
|---|---|---|
| Independent random variables X and YThe variances add for a sum and for a difference, so Var(X - Y) = Var(X) + Var(Y). The variables must be independent. X1 + X2 + X3 has variance 3 Var(X), but 3X has variance 9 Var(X). Used in: Jan 2026 Q8(a), T = 2M - L gives a = 364 and b = 5553; Jun 2024 Q6(a), the total of three bags of carrots; Jan 2025 Q7(a), 3 short and 4 long boards.Given | Given | |
| Linear combination of independent Normal variablesThe specification lists it as a formula you are expected to know. X and Y must be independent and Normal. Find the mean and the variance, then standardise. Used in: Jan 2026 (A) Q7(b), P(A > B) for task times; Jun 2023 Q7(a), P(4A - 3B < 40); Jan 2025 Q7(b), two long boards differ by more than 0.01 kg.Learn it | Learn it | |
| Total of n independent copiesUse the first for n separate observations and the second only when one value is multiplied by n. Used in: Jun 2025 Q8(a), the total of three bags of sand; Jan 2023 Q7(a), the total time for 3 customers; Jun 2022 Q6(a), 6 men and 3 women in a lift.Learn it | Learn it | |
| Expectation of a linear combinationIt needs no independence. It is how you show that an estimator is unbiased. Used in: Jun 2022 Q5(c), show that U1 and U2 are unbiased; Jun 2023 Q3(d), a + b = 1 for Y = aX1 + bX2; Jan 2026 Q8(a), the mean of 2M - L.Learn it | Learn it | |
| Quadratic formulaFrom Pure Mathematics 1. A probability condition on a total of x people or items can give a quadratic. Used in: Jun 2022 Q6(b), the maximum number of people in the lift; Jan 2025 Q7(c), the value of n from a probability.Learn it | Learn it |
Sampling
These sampling rows are formulas you must learn; the booklet prints none of them. The advantages and disadvantages of each method are in the Also learn list.
| Formula | Equation | In the exam |
|---|---|---|
| Systematic sample intervalPick a random starting point, then take every kth item from the list. N is the population and n the sample size. Used in: Jun 2025 Q1(b), 360 members and 20 players gives every 18th; Jun 2024 Q1(a), 400 employees and 8 gives every 50th; Jan 2024 Q2(a), 800 and 80 gives every 10th.Learn it | Learn it | |
| Stratified sample sizeTake each group in proportion to its size, then use random numbers within the group. Used in: Jan 2026 (A) Q6(a), 120 of 6000 full-time and 80 of 4000 part-time staff; Jan 2024 Q2(b), a sample of 100 across three offices.Learn it | Learn it |
Estimators and sampling distributions
The booklet prints the properties of the sample mean and sample variance. Bias is not printed, so learn its definition.
| Formula | Equation | In the exam |
|---|---|---|
| Sample mean as an estimator of μThe standard error of the sample mean is σ / sqrt(n), the square root of the variance. Used in: Jan 2023 Q6(d), mean of 10 large bags of gravel; Jan 2026 Q7(b), the standard error of an unbiased estimator; Jan 2025 Q4(a), show that the standard error is 0.0124.Given | Given | |
| Sample variance as an estimator of σ²Rearranged, s2 = (sum of x2 - n times the mean squared) / (n - 1), and sum of x2 = (n - 1)s2 + n times the mean squared when you combine two samples. Used in: Jan 2026 Q5(a), v = 2.3 for professor B; Jun 2022 Q2(a), unbiased estimates for brand A; Jan 2024 Q6(b)(c), unbiased estimates then a combined sample.Given | Given | |
| Bias of an estimatorWork out E(T) first using the expectation of a linear combination. Used in: Jan 2026 (A) Q2(a), the bias of the sample mean for a uniform variable; Jun 2025 Q6(b)(c), R is biased and its bias; Jan 2025 Q6(a)(b).Learn it | Learn it | |
| Uniform (rectangular) on [a, b]: meanThe mean of a continuous uniform variable on [a, b] is (a + b) / 2, so an unbiased estimator comes from rearranging E(X mean). Used in: Jan 2026 Q7(a), X mean - 2 is unbiased for a; Jan 2025 Q6(a), the mean on [2a + 3, 4a + 9]; Jun 2024 Q7(a), the mean of D.Given | Given | |
| Uniform (rectangular) on [a, b]: varianceUse (b - a)2 / 12 for the variance of one observation, then divide by n for the sample mean. Used in: Jan 2026 Q7(b), the standard error from a variance of 3; Jun 2024 Q7(a), Var(D) = 3; Jun 2022 Q4(b), the standard deviation of a wait uniform on [0, 7].Given | Given |
Confidence intervals and tests for means
The z values come from the table X3: 1.6449 for 90%, 1.9600 for 95%, 2.3263 for 98% and 2.5758 for 99%. The tables X2 and X3 follow this section.
| Formula | Equation | In the exam |
|---|---|---|
| One sample from a normal distributionUse it to test a mean with known σ, to find a critical region for the sample mean, or to set up an equation for n. Used in: Jan 2026 Q6(a), show that n = 100; Jun 2025 Q7(a), show that σ = 12; Jan 2023 Q1(c), the critical region for the sample mean.Given | Given | |
| Two independent samples from normal distributionsFor two large samples put the sample variances in place of the population variances, and take mux - muy from H0. Used in: Jan 2026 Q5(b), mean mark more than 10 higher; Jan 2026 (A) Q6(c), full-time and part-time scores; Jun 2025 Q4(b), more than 200 grams heavier.Given | Given | |
| Confidence interval for a Normal meanUse it when σ is known, or for a large sample with s in place of σ. Choose z from X3 for the level. Used in: Jan 2026 (A) Q4(a), 95% for mug masses with σ = 1.2; Jan 2023 Q6(a), 98% for stone weights; Jun 2024 Q3(b), 98% for bottle volumes.Learn it | Learn it | |
| Width and centre of an intervalHalf the width gives z σ / sqrt(n). Use it to find σ or n, rounding n up, or to rebuild an interval at a new level from an interval you are given. Used in: Jan 2026 Q3(a), a 90% interval from a 95% interval; Jun 2022 Q3(d), the smallest n for a width of at most 1.5; Jan 2025 Q4(a).Learn it | Learn it | |
| Confidence level from zFind z from the width first, then the tail probability from X2. Used in: Jun 2022 Q3(c), the value of a; Jan 2025 Q4(c), the value of r.Learn it | Learn it | |
| Intervals that contain μThe number of m independent intervals that contain μ is binomial. Count the intervals that do not contain μ with p = 1 - level. Used in: Jan 2026 Q3(b), at least 1 of 3 intervals; Jan 2026 (A) Q4(b), more than 3 of 20 do not contain μ; Jun 2024 Q3(c), more than 3 of 5.Learn it | Learn it |
Goodness of fit and contingency tables
The booklet prints the chi-squared statistic and the table X6, which follows this section. The expected frequencies and the degrees of freedom are not printed.
| Formula | Equation | In the exam |
|---|---|---|
| Goodness-of-fit test and contingency tablesThe statistic can also be written as the sum of O2 / E minus the total. Used in: Jan 2026 Q1, pass rate against test centre; Jan 2026 (A) Q5(b), a Poisson model for births; Jun 2024 Q4(a), the ratio 10 : 5 : 2 : 3 for ice cream flavours.Given | Given | |
| Expected frequency in a contingency tableCalculate it for every cell, and the expected totals must equal the observed totals. Used in: Jan 2026 (A) Q3(a), three driving test centres; Jan 2025 Q5(b), the expected frequency for Biology and London; Jan 2026 Q1.Learn it | Learn it | |
| Degrees of freedomCount the cells after combining any with E less than 5. Estimating p or λ from the data costs one more degree of freedom. Used in: Jan 2026 Q1, (2 - 1)(2 - 1) = 1; Jan 2026 Q4(e), 5 - 1 - 1 = 3 after estimating p; Jan 2025 Q2(d), 5 - 1 - 1 = 3 after estimating λ.Learn it | Learn it | |
| Expected frequencies for a fitted modelEstimate the parameter from the data unless the question gives it, then multiply each probability by the total. The expected frequencies must add to the total observed. Used in: Jan 2026 Q4(a)(b), p = 0.18 and r = 2.65; Jun 2025 Q5(a)(b), the proportion of male puppies; Jan 2025 Q2(a)(b).Learn it | Learn it | |
| Binomial B(n, p): probabilityGives the expected binomial frequencies and the probability for a number of intervals. Used in: Jan 2026 Q4(b), 150 x P(X = 5) for B(10, 0.18); Jan 2023 Q4(a), B(4, 0.5) expected frequencies; Jan 2026 Q3(b), at least 1 of 3 intervals.Given | Given | |
| Poisson Po(λ): probabilityGives the expected Poisson frequencies, which are N times P(X = x), with the last cell found by subtraction. Used in: Jan 2026 (A) Q5(a), the observed and expected frequencies for 5 births; Jan 2024 Q4(b), show that s = 8.15.Given | Given |
Correlation
Every one of the 9 papers has a Spearman's rank question. The tables X7 and X8 follow this section.
| Formula | Equation | In the exam |
|---|---|---|
| Spearman's rank correlation coefficientd is the difference between the two ranks of each pair, and n is the number of pairs. Rank both variables the same way, from the highest or from the lowest, and use average ranks for ties. Used in: Jan 2026 (A) Q1(a), distance and commission for 8 salespeople; Jan 2026 Q2(b), the largest sum of d2 for 5 dogs; Jan 2025 Q1(a).Given | Given | |
| Product moment correlation coefficientUsed when a PMCC is tested against the critical value in X7. Used in: Jan 2024 Q3(a), tea consumption and population; Jan 2023 Q2(c), r from the three S values.Given | Given |
Learn these: not in the booklet
- Sampling methodsSimple random: number the list and use random numbers. Systematic: random start then every kth, which can be biased if the list has a pattern. Stratified: proportional groups, so it reflects the population structure. Quota: needs no sampling frame but can be biased. Used in: Jun 2024 Q1(a)(b); Jan 2024 Q2(a)(c); Jan 2025 Q5(a).
- Statistic and estimatorA statistic uses only sample values and no unknown parameters, so (X - μ) / σ is not a statistic. An estimator is unbiased if its expected value equals the parameter. Used in: Jun 2022 Q5(a)(b); Jun 2023 Q3(a); Jun 2025 Q6(a).
- EfficiencyThe most efficient of several unbiased estimators has the smallest variance, found with the variance rule for the combination. Used in: Jun 2022 Q5(d).
- Central Limit TheoremFor a large sample the sample mean is approximately Normal whatever the population, and the sample variance can stand in for σ2. It is not needed when the population is already Normal. Used in: Jan 2026 (A) Q6(d)(e); Jan 2025 Q3(c); Jan 2023 Q6(b).
- Test for a meanState H0 and H1 in terms of μ, find z, compare it with the critical value from X3 (1.6449 one-tailed and 1.96 two-tailed at 5%, 2.5758 two-tailed at 1%) and conclude in context. A value of μ outside a confidence interval is rejected by the matching two-tailed test. Used in: Jan 2023 Q1(b)(c)(d); Jun 2024 Q3(a); Jan 2023 Q6(c).
- Hypotheses for a difference of meansH0 is muA - muB equal to the stated difference (0, 10 or 200) and H1 uses the sign of the claim. Used in: Jan 2026 Q5(b); Jun 2025 Q4(b); Jun 2024 Q5(a).
- Assumptions for a two-sample testSay that the samples are independent and large, so the Central Limit Theorem applies and the sample variances can be used for the population variances; mention both samples. Used in: Jan 2026 Q5(c); Jun 2024 Q5(b); Jan 2025 Q3(c).
- Chi-squared test methodState the hypotheses in words, find the expected frequencies, combine any cell with E less than 5, find the statistic and the degrees of freedom, compare with the upper-tail value in X6 and conclude in context. Used in: Jan 2026 Q4(d)(e); Jan 2025 Q2(c)(d); Jan 2026 (A) Q3(a).
- Which cell to investigateThe cell with the largest value of (O - E)2 / E contributes most to the statistic. Used in: Jan 2026 (A) Q3(b).
- Spearman's rank testRank both variables, use average ranks for ties, write the hypotheses in terms of ρ, and compare with the critical value in X8 (one-tailed or two-tailed as the question says). Used in: Jan 2026 (A) Q1(a)(b); Jun 2022 Q1(a)(c); Jan 2025 Q1(a)(b).
- Spearman or product momentUse Spearman's when the data are ranks or the relationship may not be linear; the product moment test is checked against X7. Used in: Jun 2023 Q1(a)(c); Jun 2024 Q2(c); Jan 2024 Q3(a).
- Limits on the sum of d squaredPut the critical value into the formula for the rank coefficient and solve the inequality for the sum of d2, or for the unknown in a chi-squared statistic. Used in: Jan 2026 Q2(b)(c); Jun 2022 Q7(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.
Edexcel International (IAL)1 booklet
Edexcel IAL maths units1 booklet
How to use the formula sheet in the exam
- Write down the booklet rows you need before you start: Var(aX +/- bY), the two-sample z statistic, S², the chi-squared statistic and Spearman's formula. The Normal result for aX +/- bY and the confidence interval are not printed, so learn them.
- For a confidence interval, use z from X3 for the level, then check your answer: half the width is z σ / sqrt(n) and the centre is the sample mean.
- Rank both variables the same way for Spearman's, give the hypotheses in ρ, and compare with the right column of X8. For ties, use average ranks.
- For chi-squared, combine cells with an expected frequency under 5 before you work out the degrees of freedom, and write the conclusion in the words of the question.
- Separate 'a sum of n observations' from 'n times one observation': 3 bags have variance 3 σ², and 3 times one bag has variance 9 σ².
- For two-sample tests, state that both samples are large so the Central Limit Theorem applies and the sample variances can replace the population variances.
Common questions
Which formulae does the Edexcel IAL S3 booklet give me?
The Mathematical Formulae and Statistical Tables booklet prints the variance rule for combining independent variables, the properties of the sample mean and sample variance, the one-sample and two-sample z statistics, Spearman's coefficient and the chi-squared statistic, plus the Normal tables X2 and X3, the chi-squared table X6 and the correlation tables X7 and X8. It also notes that S3 candidates may need the S1, S2 and P1 to P4 formulae.
Do I need to learn the Normal result for aX + bY?
Yes. The specification lists aX +/- bY ~ N(a mu_x +/- b mu_y, a² sigma_x² + b² sigma_y²) for independent Normal variables among the formulae you are expected to know, and says it will not appear in the booklet.
Is the confidence interval formula in the booklet?
No. The booklet prints the standardised sample mean, so you must build the interval x mean +/- z σ / sqrt(n) yourself and take z from the percentage points table X3.
Other units
Unit P1 · Unit P2 · Unit P3 · Unit P4 · Unit FP1 · Unit FP2 · Unit FP3 · Unit M1 · Unit M2 · Unit M3 · Unit S1 · Unit S2 · Unit D1. All Edexcel IAL unit 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 Edexcel IAL maths 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.