Combinations of random variables
S3 Topic 1, spec point by spec point (Combinations of random variables). Each spec point has the method, worked examples written the way the mark scheme wants them, how the marks go and the mistakes that cost the most marks.
The S3 paper (WST03) is 1 hour 30 minutes and 75 marks. It assumes S1 and S2. Most questions are a test or a confidence interval with a conclusion in context.
How maths marks work
M marks are for a correct method, even with a slip in the numbers. A marks are for accurate answers and need the M mark before them. B marks stand alone (a value or a statement). dM means the method mark depends on an earlier M mark.
ft (follow through) lets a later mark be earned from your own earlier wrong value. cso (correct solution only) means every step must be right, which is common on "show that" answers. awrt means "answers which round to", so keep extra figures and round at the end. isw means extra working after a correct answer is ignored.
When a question says "use algebra", "show all stages of your working" or "solutions relying on calculator technology are not acceptable", an answer with no method scores nothing, even if it is right.
Each spec point below has the idea behind it first, then the method, worked examples written the way a mark scheme would want, how the marks go, and the mistakes that cost the most marks. The paper references, such as (WMA11 Jan 2022 Q4), are past questions from the IAL Prep question bank that use the point.
Spec1.1Linear combinations of independent Normal variables
If \(X \sim N(\mu_x, \sigma_x^2)\) and \(Y \sim N(\mu_y, \sigma_y^2)\) are independent, then \(aX \pm bY \sim N(a\mu_x \pm b\mu_y,\ a^2\sigma_x^2 + b^2\sigma_y^2)\). No proofs required.
The three facts you need
- Means combine as you'd expect: \(E(aX \pm bY) = aE(X) \pm bE(Y)\).
- Variances always add (for independent variables), and coefficients are squared:
\[\text{Var}(aX \pm bY) = a^2\text{Var}(X) + b^2\text{Var}(Y)\]
- A combination of independent Normals is still Normal, so once you have the mean and variance you standardise as usual.
Why variances add even for a difference: subtracting a variable doesn't cancel its randomness; it adds more uncertainty. And \((-b)^2 = b^2\).
\(X_1 + X_2\) is not the same as \(2X\)
This is the idea that most S3 marks depend on.
| Situation | Mean | Variance |
|---|---|---|
| Two separate items, \(X_1 + X_2\) | \(2\mu\) | \(\sigma^2 + \sigma^2 = 2\sigma^2\) |
| One item doubled, \(2X\) | \(2\mu\) | \(2^2\sigma^2 = 4\sigma^2\) |
| Total of \(n\) separate items | \(n\mu\) | \(n\sigma^2\) |
| One item multiplied by \(n\) | \(n\mu\) | \(n^2\sigma^2\) |
Read the wording: "the total mass of 3 bags" is \(X_1 + X_2 + X_3\); "three times the mass of a bag" is \(3X\).
Comparing two quantities
To find \(P(A > B)\), form \(D = A - B\) and find \(P(D > 0)\). This works for "the large bag is heavier than three small bags", "the man is more than twice as heavy as the woman", and so on.
a total
Task times are independent with \(A \sim N(21, 2^2)\), \(B \sim N(32, 7^2)\), \(C \sim N(45, 9^2)\). Find the probability the total time exceeds 90 minutes.
\(T = A + B + C\): \(E(T) = 98\), \(\text{Var}(T) = 4 + 49 + 81 = 134\).
\(P(T > 90) = P\left(Z > \frac{90 - 98}{\sqrt{134}}\right) = P(Z > -0.691) = 0.755\).
a lift
Men's masses are \(N(80, 10^2)\) and women's \(N(69, 5^2)\). Find the probability that 6 men and 3 women exceed the lift limit of 700 kg.
Mean \(= 6(80) + 3(69) = 687\). Variance \(= 6(100) + 3(25) = 675\) (separate people, so \(6\sigma^2\), not \(36\sigma^2\)).
\(P(\text{total} > 700) = P\left(Z > \frac{13}{\sqrt{675}}\right) = P(Z > 0.50) = 0.308\).
one item against several
Small bags \(F \sim N(1.00, 0.02^2)\) kg, large bags \(L \sim N(3.05, 0.04^2)\) kg. Find the probability a large bag weighs more than three small bags.
\(D = L - (F_1 + F_2 + F_3)\): \(E(D) = 0.05\), \(\text{Var}(D) = 0.04^2 + 3(0.02^2) = 0.0028\).
\(P(D > 0) = P\left(Z > \frac{-0.05}{\sqrt{0.0028}}\right) = P(Z > -0.945) = 0.828\).
(Using \(L - 3F\) instead would give variance \(0.0016 + 9(0.0004) = 0.0052\) and the wrong answer, 0.756.)
- M1 for a fully correct method for the mean or variance, A1 for both, M1 for standardising with their values, A1 awrt the probability (WST03A Jan 2026 Q7a).
- The lift question: B1 for the mean 687, B1 for the variance 675, M1 standardise, A1 (WST03 Jun 2022 Q6a).
- Harder versions give a probability and ask for \(n\): e.g. \(Y = (n - 1)T_1 - (T_2 + \ldots + T_n)\) has variance \((n - 1)^2\sigma^2 + (n - 1)\sigma^2\); the coefficient of \(T_1\) is squared but the others are 1. Standardise, set equal to the \(z\) value, and solve a quadratic in \(n\) (WST03 Oct 2020 Q7c).
Often asked as: by far the most examined S3 point (50 questions). Totals; differences; "more than twice"; finding an unknown mean, variance or \(n\) from a given probability.
\(\text{Var}(X - Y) = \text{Var}(X) - \text{Var}(Y)\).
Using \(n^2\sigma^2\) for a total of \(n\) separate items (or \(n\sigma^2\) for one item scaled by \(n\)).
Forgetting to square-root the variance before standardising.