Loav · VCE Specialist Mathematics · Units 3 & 4
S34-S01 · Recommended sequence: Unit 4
1. Decide what is random before doing arithmetic
A linear combination adds scaled random variables, possibly with a fixed constant. If X is a parcel's mass and Y is its packaging mass, then X + Y is a total mass. If X and Y are completion times for two processes, X − Y measures their difference. The algebra looks familiar, but a random variable represents a collection of possible outcomes rather than one observed number. Its mean describes the centre of that collection and its variance describes spread around that centre.
Use μX = E(X), σX² = Var(X) and σX = SD(X). Variance has squared measurement units; standard deviation has the original units. Throughout these guides, N(μ, σ²) uses the variance as its second parameter. Many calculator commands instead ask for standard deviation, so translating the notation correctly is an essential step, not a cosmetic choice.
A fixed constant is not a random variable with unexplained uncertainty. Adding a known 5 g label shifts every possible mass upward by exactly 5 g. It changes the mean but leaves all distances from the mean unchanged. Scaling by a factor changes those distances as well as the centre. A negative scale reverses their signs, but distances squared remain nonnegative.
Var(aX + b) = a²Var(X)
SD(aX + b) = |a|SD(X)
The absolute value in the standard-deviation rule matters. For Y = −3X + 7, the mean is −3μX + 7, variance is 9σX², and standard deviation is 3σX. A negative standard deviation cannot describe a spread.
2. Means add; independent variation adds in quadrature
Expectation is linear: E(aX + bY + c) = aE(X) + bE(Y) + c. This mean rule does not itself require independence. To use the corresponding variance rule in this course, however, the variables must be independent. Independence means knowing one variable's outcome does not change the probability distribution of the other. It is stronger than simply having different variable names or measuring different objects.
Var(aX + bY + c) = a²σX² + b²σY²
Notice that a difference still has a sum of variance contributions. In X − Y, an unusually large X can make the difference large, while an unusually large Y can make it small. Both introduce uncertainty. Subtracting the variables does not cancel their independent fluctuations. The coefficient −1 becomes +1 when squared.
Standard deviations generally do not add. First convert each standard deviation to a variance, apply the squared coefficients, add, and only then take a square root if a standard deviation is required. The result is a square-root-of-squares calculation, rather like combining perpendicular components. This analogy is a memory aid, not evidence that independence holds in a particular context.
When independence is not supplied or defensible, stop before using variance additivity. Shared temperature, common calibration or repeated use of the same observation can make it unsuitable. The mean may still be calculable from linearity, but the provided information may not determine the variance. Do not invent an independence assumption silently.
Worked example 1: A weighted output
A production score is S = 2X − 3Y + 10. Independent variables X and Y have means 8 and 5, and standard deviations 1.5 and 2 respectively. Find the mean, variance and standard deviation of S. No distributional shape is given.
The mean is E(S) = 2(8) − 3(5) + 10 = 11. For the variance, square both coefficients and both supplied standard deviations: Var(S) = 2²(1.5²) + (−3)²(2²) = 9 + 36 = 45. Therefore SD(S) = √45 = 3√5 ≈ 6.708.
The constant 10 has contributed to the mean only. The negative coefficient of Y has produced a positive variance contribution. These moments are valid without assuming normality, but they do not establish the shape of S's distribution. It would be unjustified to enter 11 and √45 into a normal probability command merely because those two numbers are available.
3. A total of fresh observations is different from a multiple
Let X₁, X₂, …, Xₙ be independent and identically distributed, abbreviated iid. They share the same mean μ and variance σ², but each represents a fresh random outcome. Their total T = X₁ + ··· + Xₙ has E(T) = nμ and Var(T) = nσ². Consequently SD(T) = √n σ.
By contrast, nX means one random outcome multiplied by n. Its mean is also nμ, but its variance is n²σ² and standard deviation is nσ. There are no fresh observations whose high and low deviations can partially balance. Every repeated contribution is the very same value. Equal means therefore do not guarantee equal distributions.
Suppose one packet has mean mass 10 g and standard deviation 2 g. Three independently selected packets have total mean 30 g and standard deviation √12 g. Triple the mass of one selected packet and the mean remains 30 g, while the standard deviation becomes 6 g. A question about three packets should usually use three variables; a question about a scale displaying three times one reading should use 3X.
4. Normal combinations give a complete probability model
A linear combination of independent normal random variables is itself normal. Calculate its mean and variance first, then write the new distribution explicitly. If a nonzero combination W has mean m and standard deviation s, standardise with Z = (W − m)/s. This separates the modelling decision from the normal-area calculation.
For a right-tail probability, use 1 − Φ(z); for a left tail use Φ(z); for an interval subtract cumulative probabilities. Here Φ denotes the cumulative distribution function of N(0, 1). A sketch should show whether the required area is small or large. Negative coefficients do not prevent normality. A completely constant combination, with zero variance, is handled directly rather than by division by zero.
Worked example 2: A clearance rather than a total
A manufactured opening has width X ∼ N(12.5, 0.6²) mm and an independently produced insert has width Y ∼ N(10, 0.4²) mm. Find the probability that the clearance X − Y is less than 2 mm.
Define D = X − Y. Independence gives E(D) = 12.5 − 10 = 2.5 mm and Var(D) = 0.6² + 0.4² = 0.52 mm². Normality of both inputs gives D ∼ N(2.5, 0.52), with standard deviation √0.52 ≈ 0.7211 mm.
Therefore Pr(D < 2) = Φ[(2 − 2.5)/√0.52] = Φ(−0.6934…) ≈ 0.2440. The answer being below one half is sensible: 2 mm lies below the mean clearance. This probability includes negative clearances, which represent an insert wider than the opening. If a question instead conditions on a positive clearance, a conditional probability calculation is needed; that restriction cannot be added without changing the event.
Worked example 3: Independent packages in a crate
Sixteen independent packages have masses distributed as N(92.5, 10²) g. A fixed empty crate has mass 520 g. Find the chance that the loaded crate exceeds 2020 g.
Let T be the total package mass. Then T ∼ N(16 × 92.5, 16 × 100) = N(1480, 1600). Its standard deviation is 40 g, not 160 g. The loaded mass L = T + 520 consequently has distribution N(2000, 1600).
Thus Pr(L > 2020) = Pr[T > 1500] = 1 − Φ[(1500 − 1480)/40] = 1 − Φ(0.5) ≈ 0.3085. Both routes give the same event and provide a useful check. The fixed crate contributes no variance. If its mass varied independently between crates, that extra variance would need to be included rather than hidden inside the constant.
5. A non-routine choice and a technology check
Two independent measurements of the same quantity have common mean μ and variances 4 and 9. A weighted estimate W = aX + (1 − a)Y retains mean μ for every real a. Its variance is 4a² + 9(1 − a)² = 13(a − 9/13)² + 36/13. Hence the smallest variance occurs at a = 9/13, giving greater weight to the more precise measurement. This is an application of a quadratic, not a requirement to learn a new estimation theory.
To check the calculation numerically, choose a particular μ, generate independent normal values with standard deviations 2 and 3, and calculate W for many pairs. Compare the simulated mean and variance with μ and 36/13. Keep the two random draws separate. Reusing the same random draw for X and Y changes the experiment and invalidates the independence calculation.
Technology should check a formulated model. Record whether an input field requests variance or standard deviation, use unrounded intermediate values, and retain a visible probability statement. A normal calculator cannot decide whether package masses are independent or whether a fixed mass was accidentally counted sixteen times. Those are mathematical modelling decisions.
Practice questions
- X has mean 7 and variance 9. Find the mean and standard deviation of 4 − 2X.
- Independent X and Y have means 3 and 5 and variances 4 and 9. Find the mean and variance of 3X − 2Y + 1.
- Five iid observations each have mean 6 and standard deviation 2. Compare their total with five times one observation.
- Independent X ∼ N(12, 4) and Y ∼ N(4, 1). Find Pr(X − Y < 10).
- A fixed container weighs 100 g. Nine independent contents each have mean 20 g and variance 16 g². Find the loaded mean and standard deviation.
- Independent normal measurements X and Y each have mean μ and variance 16. Find the distribution of (X + Y)/2 and explain its advantage over one measurement.
- A student knows E(X) = 10, E(Y) = 4 and both variances are 9, but has no information about dependence. Which of E(X − Y) and Var(X − Y) can be found from these facts?
Fully reasoned answers
- Linearity gives 4 − 2(7) = −10. The variance is (−2)² × 9 = 36, so the standard deviation is 6. Subtracting from 4 reverses and shifts the distribution but cannot make its spread negative.
- The mean is 3(3) − 2(5) + 1 = 0. Independence allows variance additivity after squaring coefficients: 9(4) + 4(9) = 72. The constant contributes zero variance; a zero mean does not mean a constant score.
- The independent total has mean 30, variance 5(4) = 20 and standard deviation 2√5. Five times one observation has mean 30, variance 25(4) = 100 and standard deviation 10. The distinction is fresh observations versus one repeated outcome.
- D = X − Y is normal with mean 8 and variance 4 + 1 = 5. Therefore Pr(D < 10) = Φ(2/√5) ≈ 0.8145. The second parameters supplied were variances, so they must not be squared again.
- The loaded mean is 100 + 9(20) = 280 g. Its variance is 9(16) = 144 g² and standard deviation is 12 g. The constant container mass shifts the total without increasing its uncertainty.
- The mean is μ and the variance is (1/2)²(16) + (1/2)²(16) = 8. Thus the average is N(μ, 8), with standard deviation 2√2 rather than 4. It preserves the centre while reducing random variation under the independence model.
- E(X − Y) = 10 − 4 = 6 follows without independence. The variance is not determined by the supplied facts: using 9 + 9 would need independence or additional suitable information. Equal individual variances do not justify that assumption.
Curriculum and sources
VCE Specialist Mathematics Units 3 and 4, Areas of Study 6 as applicable, with Outcomes 1–3. Scope follows the VCAA Mathematics Study Design 2023–2027 and the current Specialist Mathematics FAQ. Curriculum checked 15 October 2026. Teaching reference: Cambridge Senior Mathematics VCE Specialist Mathematics 3&4, second edition (2023), sections 15A–15C (pp646–662). All teaching prose, examples, practice questions and diagrams are original. Independent teaching material; not a VCAA publication.
