JC2 H2 Mathematics Tutor Bukit Timah | Why Variances Add When Values Subtract

JC2 H2 Mathematics tutor Bukit Timah — Did you know? Subtracting two independent random quantities does not mean subtracting their variances. The difference may be smaller on average, yet its uncertainty is the sum of the two uncertainties. If independent X and Y have variances 9 and 4, then X−Y has variance 13, not 5. This can seem odd until you imagine each quantity fluctuating independently: a high X and a low Y can make the difference exceptionally large in magnitude.

This Singapore–Cambridge JC2 H2 Mathematics 9758 guide is for Bukit Timah parents, tutors and JC2 learners who want a quick way to choose the correct formula for E(aX+bY) and Var(aX+bY). We will start with two independent random variables, scale them by constants, add and subtract them, examine transformed normal variables and practise interpreting results in fictional contexts. The 2027 H2 syllabus includes these combinations when the variables are independent. No invented scenario here is evidence about actual learners, workers or businesses.

Read the quick answer first, then the worked calculation, one-minute decision test and 56 original clinics. The central judgement is whether your question concerns the centre of a distribution or its spread. Most mistaken answers come from applying mean rules to variance, forgetting to square a coefficient, or assuming independence without being told. The official curriculum reference is the 2027 H2 Mathematics syllabus 9758.

Quick answer: linear means; squared-coefficient variances

For random variables X and Y with existing expectations, E(aX+bY)=aE(X)+bE(Y) for any fixed numbers a and b. This expectation rule does not require independence. If X and Y are independent and have finite variances, the variance formula simplifies to Var(aX+bY)=a²Var(X)+b²Var(Y). The coefficients are squared and the two contributions are added. Subtracting Y simply means b is negative, and b² is still positive.

• Independent X+Y: E(X+Y)=μX+μY and Var(X+Y)=σX²+σY².
• Independent X−Y: E(X−Y)=μX−μY and Var(X−Y)=σX²+σY².
• Scaled variable: E(3X)=3μX; Var(3X)=9σX².
• Shifted variable: E(X+7)=μX+7; Var(X+7)=σX².
• Standard deviation: take the positive square root of the variance after combining independent contributions.
• Independence warning: without independence, do not automatically use the simple variance-addition rule.

A memorable example: the difference still fluctuates

Imagine two independently measured fictional delivery intervals. The first fluctuates around 12 minutes and the second around 8 minutes. Their expected difference is four minutes. But that difference is not perfectly predictable: the first interval may run long while the second runs short, or vice versa. The difference can therefore vary even though the expected times partially cancel. Subtracting average values changes the centre; it does not remove independent sources of uncertainty.

An even simpler mental experiment uses X and Y, independently taking values +1 and −1 with equal probabilities. Both have mean zero and variance one. The difference X−Y can be −2, 0 or +2; its mean is zero but its variance is two. The outcome −2 occurs when X=−1 and Y=+1, while +2 occurs when X=+1 and Y=−1. This small enumeration makes it clear why a variance calculation of 1−1=0 would be absurd: it would claim the difference never varies at all.

The complete worked example: X−Y and 2X−3Y

Assume X and Y are independent, with E(X)=12, Var(X)=9, E(Y)=8 and Var(Y)=4. Define D = X−Y. Expectation behaves linearly: E(D)=12−8=4. Variance, however, uses squared coefficients: Var(D)=1²(9)+(−1)²(4)=13. The standard deviation is SD(D)=√13≈3.6056. Do not call 13 a standard deviation: it is a variance and has squared units.

For the sum S=X+Y, E(S)=12+8=20 while Var(S)=9+4=13. The sum and difference have the same variance here because the only change is the sign of a coefficient whose square remains one. They do not have the same mean. This illustrates why mean and variance must be handled on separate lines of working, with separate explanations.

Now define W=2X−3Y. Its expected value is E(W)=2(12)−3(8)=0. Its variance is Var(W)=2²(9)+(−3)²(4)=36+36=72, giving standard deviation √72≈8.4853. W has a zero mean but clearly not zero variance. The large spread is possible because both independent inputs have been magnified by their coefficients. An answer of 2×9−3×4=6 would misuse the expectation rule on squared deviations.

A comparison table worth remembering

Random expression	Expectation	Variance, given independence
X+Y	μX+μY	σX²+σY²
X−Y	μX−μY	σX²+σY²
2X+Y	2μX+μY	4σX²+σY²
2X−3Y	2μX−3μY	4σX²+9σY²
−X+Y	−μX+μY	σX²+σY²
X+5	μX+5	σX²
−2X+5	−2μX+5	4σX²

Look at the minus signs in the expectation column and the squares in the variance column. The signs matter for the expected value, but scaling a random fluctuation up or down by the same magnitude has the same effect on its variance. If you replace X by −X, the distribution reflects around zero, its mean changes sign, and its variance does not change. That is a geometric fact, not a special trick of the calculator.

Why coefficients are squared in variance

Variance measures the mean of squared deviations from the mean. Let U = aX+b for fixed a and b. Its expectation is aE(X)+b, so the deviation U−E(U) equals a[X−E(X)]; the constant b disappears. Squaring that deviation produces a²[X−E(X)]². Taking expectations gives Var(aX+b)=a²Var(X). This argument does not need independence because only one random variable appears.

If X has standard deviation 4, then 3X has standard deviation 12 and variance 144, not 12 and not 48. Multiplying observations by three multiplies every distance from the mean by three. Squaring those distances multiplies variance by nine. A learner who can explain this physically is less likely to forget the square in later work with combinations.

Why independent variance contributions add

For a sum X+Y, the deviations from the combined mean are [X−E(X)]+[Y−E(Y)]. Their square has individual squared terms and a mixed product. Under independence and finite-variance assumptions, that mixed product has expectation zero, leaving Var(X)+Var(Y). For a difference, the mixed term has the opposite sign, but it still has expectation zero under independence, leaving the same addition of the individual variances.

This shows exactly where independence enters. Mean linearity is general, but the simple variance-addition rule relies on the relationship between the random variables. In a real context, two measurements can share a common influence. If the question does not grant independence or other appropriate information, a student should not invent it merely to obtain a neat numerical answer. The 2027 syllabus explicitly frames E(aX+bY) and Var(aX+bY) for the case where X and Y are independent.

The normal-distribution connection

If X and Y are both normally distributed and independent, a linear combination aX+bY is also normal. In the worked example, if X~N(12,9) and Y~N(8,4) independently, then D=X−Y follows N(4,13). Notice that the second parameter in the normal notation is the variance 13, while the standard deviation is √13. To compute a probability such as P(D>7), standardise using (7−4)/√13≈0.83205, not divide by 13.

If X and Y are merely independent with the listed means and variances but no normality assumption, their linear combination still has the calculated mean and variance. You cannot automatically conclude its entire distribution is normal. This distinction between knowing two moments and knowing the full probability model is crucial. A mean and variance do not uniquely identify an arbitrary distribution.

A unit check that detects absurd answers

Suppose X and Y both measure time in minutes. E(X−Y) is measured in minutes, Var(X−Y) in minutes squared, and SD(X−Y) in minutes. If W=2X−3Y, the coefficients are dimensionless multipliers and those unit statements remain the same. A final answer such as “variance = 72 minutes” is dimensionally inconsistent: it should be “72 minutes²” under the fictional measurement model. Units are a cheap and powerful verification tool.

If quantities are measured in incompatible units, the expression aX+bY requires coefficients that make the terms commensurate. Do not add a distance in metres to a time in seconds without a meaningful conversion or model coefficient. The algebra of expectations can be performed symbolically, but the result only has a usable real-world interpretation when the quantities and their units have been defined.

What if X and Y are actually the same variable?

Consider X with nonzero variance. If someone writes X−X, the result is identically zero and has variance zero. A careless use of the independent rule would produce Var(X)+Var(X), falsely implying nonzero spread. The issue is that the two X terms are not independent variables at all — they are precisely the same random quantity. This is the simplest counterexample to the habit of adding variances without checking the problem’s assumptions.

The independence statement is not decoration. If the question explicitly calls the inputs independent, the simple squared-coefficient rule applies. If it states they are linked, or gives no relationship, pause before concluding. A school problem may be designed to reward the observation that the information is insufficient. Giving an unjustified exact numerical variance is not mathematical confidence; it is an unstated assumption.

The one-minute classroom diagnostic

Ask the learner for E(X−Y) and Var(X−Y), given independent Var(X)=9, Var(Y)=4 and means 12, 8. They should answer 4 and 13 in that order. Next replace X−Y by 2X−3Y. They should obtain 0 and 72. Then erase the word “independent” and ask whether 72 is still guaranteed. A strong learner will say no, while noticing that the expected value remains 0 from linearity.

Finally ask about X−X. Its variance is exactly zero, not 18. That changed-case question is often more revealing than another ten substitutions because it tests whether the learner has understood the hypothesis required for variance addition. If they can explain the counterexample without referring to a formula sheet, the conceptual repair has succeeded.

Fifty-six original workshops on independent means and variances

Every workshop is an original fictional exercise with stated independent inputs. Some describe independent normal measurements, but most need only means and variances; they do not invent an entire distribution from those two moments. Work with the coefficients first, then the numeric arithmetic. A tutor can hide the worked answer and ask the pupil to write the formula in symbols, the computation and a one-sentence interpretation. The values vary in scale and sign so a learner cannot succeed by memorising the result of the first example.

1. Independent time readings — coefficients 1, -2

Invented setup. Let independent X and Y represent two independent time readings with means 6 and 4 and standard deviations 1 and 4. A derived score is W=1X − 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 4²=16. The expected value is E(W)=(1)(6)+(-2)(4)=-2. Because X and Y are independent, Var(W)=(1)²(1)+(-2)²(16)=65. The standard deviation is √65≈8.06226. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. When the total expected value is negative, the result remains perfectly legitimate; an expectation can lie below zero even when both variance contributions are non-negative. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 65 but the expected value becomes 14. Explain why this changed case is not a paradox. A result that uses subtracted variance 64 with a negative sign would be confusing an operation on values with an operation on squared deviations.

2. Separate batch weights — coefficients 2, 2

Invented setup. Let independent X and Y represent two separate batch weights with means 13 and 9 and standard deviations 2 and 5. A derived score is W=2X + 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 5²=25. The expected value is E(W)=(2)(13)+(2)(9)=44. Because X and Y are independent, Var(W)=(2)²(4)+(2)²(25)=116. The standard deviation is √116≈10.77033. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. A coefficient of −3 affects the mean through its negative sign but affects variance through the positive factor 9. Name that distinction explicitly. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 116 but the expected value becomes 8. Explain why this changed case is not a paradox. A result that uses added variance 100 with a negative sign would be confusing an operation on values with an operation on squared deviations.

3. Two independent trial scores — coefficients -1, -4

Invented setup. Let independent X and Y represent two two independent trial scores with means 20 and 14 and standard deviations 3 and 6. A derived score is W=-1X − 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 6²=36. The expected value is E(W)=(-1)(20)+(-4)(14)=-76. Because X and Y are independent, Var(W)=(-1)²(9)+(-4)²(36)=585. The standard deviation is √585≈24.18677. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Reverse the sign of the second coefficient and predict which of the two outputs must change. The variance will be the same; the expectation generally will not. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 585 but the expected value becomes 36. Explain why this changed case is not a paradox. A result that uses subtracted variance 576 with a negative sign would be confusing an operation on values with an operation on squared deviations.

4. Independent fictional demand levels — coefficients 3, 3

Invented setup. Let independent X and Y represent two independent fictional demand levels with means 27 and 19 and standard deviations 4 and 2. A derived score is W=3X + 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 2²=4. The expected value is E(W)=(3)(27)+(3)(19)=138. Because X and Y are independent, Var(W)=(3)²(16)+(3)²(4)=180. The standard deviation is √180≈13.41641. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. If a friend has added the standard deviations rather than the variances, square the proposed answer and compare it with the computed variance. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 180 but the expected value becomes 24. Explain why this changed case is not a paradox. A result that uses added variance 36 with a negative sign would be confusing an operation on values with an operation on squared deviations.

5. Independent temperature deviations — coefficients -2, -3

Invented setup. Let independent X and Y represent two independent temperature deviations with means 11 and 5 and standard deviations 5 and 3. A derived score is W=-2X − 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 3²=9. The expected value is E(W)=(-2)(11)+(-3)(5)=-37. Because X and Y are independent, Var(W)=(-2)²(25)+(-3)²(9)=181. The standard deviation is √181≈13.45362. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. This exercise gives only the first two moments. Unless distributions are specified, do not announce that the linear combination is normal. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 181 but the expected value becomes -7. Explain why this changed case is not a paradox. A result that uses subtracted variance 81 with a negative sign would be confusing an operation on values with an operation on squared deviations.

6. Two independent processing delays — coefficients -3, 1

Invented setup. Let independent X and Y represent two two independent processing delays with means 18 and 10 and standard deviations 6 and 4. A derived score is W=-3X + 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 4²=16. The expected value is E(W)=(-3)(18)+(1)(10)=-44. Because X and Y are independent, Var(W)=(-3)²(36)+(1)²(16)=340. The standard deviation is √340≈18.43909. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The coefficients may represent unit conversions or scaled quantities; confirm that both terms have compatible units before interpreting a combined outcome. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 340 but the expected value becomes -64. Explain why this changed case is not a paradox. A result that uses added variance 16 with a negative sign would be confusing an operation on values with an operation on squared deviations.

7. Independent component measurements — coefficients 4, -1

Invented setup. Let independent X and Y represent two independent component measurements with means 25 and 15 and standard deviations 1 and 5. A derived score is W=4X − 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 5²=25. The expected value is E(W)=(4)(25)+(-1)(15)=85. Because X and Y are independent, Var(W)=(4)²(1)+(-1)²(25)=41. The standard deviation is √41≈6.40312. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try changing the first coefficient from a to 2a. Its contribution to variance quadruples; its contribution to the mean doubles. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 41 but the expected value becomes 115. Explain why this changed case is not a paradox. A result that uses subtracted variance 25 with a negative sign would be confusing an operation on values with an operation on squared deviations.

8. Independent route variations — coefficients -4, 4

Invented setup. Let independent X and Y represent two independent route variations with means 9 and 20 and standard deviations 2 and 6. A derived score is W=-4X + 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 6²=36. The expected value is E(W)=(-4)(9)+(4)(20)=44. Because X and Y are independent, Var(W)=(-4)²(4)+(4)²(36)=640. The standard deviation is √640≈25.29822. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try adding a fixed constant 5 to the whole expression. Its mean increases by 5 but its variance remains identical. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 640 but the expected value becomes -116. Explain why this changed case is not a paradox. A result that uses added variance 576 with a negative sign would be confusing an operation on values with an operation on squared deviations.

9. Two independent calibration errors — coefficients 1, -2

Invented setup. Let independent X and Y represent two two independent calibration errors with means 16 and 6 and standard deviations 3 and 2. A derived score is W=1X − 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 2²=4. The expected value is E(W)=(1)(16)+(-2)(6)=4. Because X and Y are independent, Var(W)=(1)²(9)+(-2)²(4)=25. The standard deviation is √25≈5. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. An expected result of zero does not imply a certain result of zero. Positive variance records the remaining fluctuation. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 25 but the expected value becomes 28. Explain why this changed case is not a paradox. A result that uses subtracted variance 16 with a negative sign would be confusing an operation on values with an operation on squared deviations.

10. Independent inventory counts — coefficients 2, 2

Invented setup. Let independent X and Y represent two independent inventory counts with means 23 and 11 and standard deviations 4 and 3. A derived score is W=2X + 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 3²=9. The expected value is E(W)=(2)(23)+(2)(11)=68. Because X and Y are independent, Var(W)=(2)²(16)+(2)²(9)=100. The standard deviation is √100≈10. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Do not confuse the variance of a combination with the variance of a sample mean; averaging multiple measurements is a related but distinct construction. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 100 but the expected value becomes 24. Explain why this changed case is not a paradox. A result that uses added variance 36 with a negative sign would be confusing an operation on values with an operation on squared deviations.

11. Separate sensor outputs — coefficients -1, -4

Invented setup. Let independent X and Y represent two separate sensor outputs with means 7 and 16 and standard deviations 5 and 4. A derived score is W=-1X − 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 4²=16. The expected value is E(W)=(-1)(7)+(-4)(16)=-71. Because X and Y are independent, Var(W)=(-1)²(25)+(-4)²(16)=281. The standard deviation is √281≈16.76305. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Suppose the two measurements shared a common random source. The simple addition of variances would then need justification, not automatic use. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 281 but the expected value becomes 57. Explain why this changed case is not a paradox. A result that uses subtracted variance 256 with a negative sign would be confusing an operation on values with an operation on squared deviations.

12. Independent modelled amounts — coefficients 3, 3

Invented setup. Let independent X and Y represent two independent modelled amounts with means 14 and 21 and standard deviations 6 and 5. A derived score is W=3X + 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 5²=25. The expected value is E(W)=(3)(14)+(3)(21)=105. Because X and Y are independent, Var(W)=(3)²(36)+(3)²(25)=549. The standard deviation is √549≈23.43075. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The final standard deviation is the positive square root of the variance, not the signed square root suggested by the original subtraction symbol. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 549 but the expected value becomes -21. Explain why this changed case is not a paradox. A result that uses added variance 225 with a negative sign would be confusing an operation on values with an operation on squared deviations.

13. Independent time readings — coefficients -2, -3

Invented setup. Let independent X and Y represent two independent time readings with means 21 and 7 and standard deviations 1 and 6. A derived score is W=-2X − 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 6²=36. The expected value is E(W)=(-2)(21)+(-3)(7)=-63. Because X and Y are independent, Var(W)=(-2)²(1)+(-3)²(36)=328. The standard deviation is √328≈18.11077. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. When the total expected value is negative, the result remains perfectly legitimate; an expectation can lie below zero even when both variance contributions are non-negative. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 328 but the expected value becomes -21. Explain why this changed case is not a paradox. A result that uses subtracted variance 324 with a negative sign would be confusing an operation on values with an operation on squared deviations.

14. Separate batch weights — coefficients -3, 1

Invented setup. Let independent X and Y represent two separate batch weights with means 28 and 12 and standard deviations 2 and 2. A derived score is W=-3X + 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 2²=4. The expected value is E(W)=(-3)(28)+(1)(12)=-72. Because X and Y are independent, Var(W)=(-3)²(4)+(1)²(4)=40. The standard deviation is √40≈6.32456. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. A coefficient of −3 affects the mean through its negative sign but affects variance through the positive factor 9. Name that distinction explicitly. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 40 but the expected value becomes -96. Explain why this changed case is not a paradox. A result that uses added variance 4 with a negative sign would be confusing an operation on values with an operation on squared deviations.

15. Two independent trial scores — coefficients 4, -1

Invented setup. Let independent X and Y represent two two independent trial scores with means 12 and 17 and standard deviations 3 and 3. A derived score is W=4X − 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 3²=9. The expected value is E(W)=(4)(12)+(-1)(17)=31. Because X and Y are independent, Var(W)=(4)²(9)+(-1)²(9)=153. The standard deviation is √153≈12.36932. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Reverse the sign of the second coefficient and predict which of the two outputs must change. The variance will be the same; the expectation generally will not. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 153 but the expected value becomes 65. Explain why this changed case is not a paradox. A result that uses subtracted variance 9 with a negative sign would be confusing an operation on values with an operation on squared deviations.

16. Independent fictional demand levels — coefficients -4, 4

Invented setup. Let independent X and Y represent two independent fictional demand levels with means 19 and 22 and standard deviations 4 and 4. A derived score is W=-4X + 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 4²=16. The expected value is E(W)=(-4)(19)+(4)(22)=12. Because X and Y are independent, Var(W)=(-4)²(16)+(4)²(16)=512. The standard deviation is √512≈22.62742. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. If a friend has added the standard deviations rather than the variances, square the proposed answer and compare it with the computed variance. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 512 but the expected value becomes -164. Explain why this changed case is not a paradox. A result that uses added variance 256 with a negative sign would be confusing an operation on values with an operation on squared deviations.

17. Independent temperature deviations — coefficients 1, -2

Invented setup. Let independent X and Y represent two independent temperature deviations with means 26 and 8 and standard deviations 5 and 5. A derived score is W=1X − 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 5²=25. The expected value is E(W)=(1)(26)+(-2)(8)=10. Because X and Y are independent, Var(W)=(1)²(25)+(-2)²(25)=125. The standard deviation is √125≈11.18034. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. This exercise gives only the first two moments. Unless distributions are specified, do not announce that the linear combination is normal. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 125 but the expected value becomes 42. Explain why this changed case is not a paradox. A result that uses subtracted variance 100 with a negative sign would be confusing an operation on values with an operation on squared deviations.

18. Two independent processing delays — coefficients 2, 2

Invented setup. Let independent X and Y represent two two independent processing delays with means 10 and 13 and standard deviations 6 and 6. A derived score is W=2X + 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 6²=36. The expected value is E(W)=(2)(10)+(2)(13)=46. Because X and Y are independent, Var(W)=(2)²(36)+(2)²(36)=288. The standard deviation is √288≈16.97056. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The coefficients may represent unit conversions or scaled quantities; confirm that both terms have compatible units before interpreting a combined outcome. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 288 but the expected value becomes -6. Explain why this changed case is not a paradox. A result that uses added variance 144 with a negative sign would be confusing an operation on values with an operation on squared deviations.

19. Independent component measurements — coefficients -1, -4

Invented setup. Let independent X and Y represent two independent component measurements with means 17 and 18 and standard deviations 1 and 2. A derived score is W=-1X − 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 2²=4. The expected value is E(W)=(-1)(17)+(-4)(18)=-89. Because X and Y are independent, Var(W)=(-1)²(1)+(-4)²(4)=65. The standard deviation is √65≈8.06226. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try changing the first coefficient from a to 2a. Its contribution to variance quadruples; its contribution to the mean doubles. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 65 but the expected value becomes 55. Explain why this changed case is not a paradox. A result that uses subtracted variance 64 with a negative sign would be confusing an operation on values with an operation on squared deviations.

20. Independent route variations — coefficients 3, 3

Invented setup. Let independent X and Y represent two independent route variations with means 24 and 4 and standard deviations 2 and 3. A derived score is W=3X + 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 3²=9. The expected value is E(W)=(3)(24)+(3)(4)=84. Because X and Y are independent, Var(W)=(3)²(4)+(3)²(9)=117. The standard deviation is √117≈10.81665. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try adding a fixed constant 5 to the whole expression. Its mean increases by 5 but its variance remains identical. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 117 but the expected value becomes 60. Explain why this changed case is not a paradox. A result that uses added variance 81 with a negative sign would be confusing an operation on values with an operation on squared deviations.

21. Two independent calibration errors — coefficients -2, -3

Invented setup. Let independent X and Y represent two two independent calibration errors with means 8 and 9 and standard deviations 3 and 4. A derived score is W=-2X − 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 4²=16. The expected value is E(W)=(-2)(8)+(-3)(9)=-43. Because X and Y are independent, Var(W)=(-2)²(9)+(-3)²(16)=180. The standard deviation is √180≈13.41641. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. An expected result of zero does not imply a certain result of zero. Positive variance records the remaining fluctuation. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 180 but the expected value becomes 11. Explain why this changed case is not a paradox. A result that uses subtracted variance 144 with a negative sign would be confusing an operation on values with an operation on squared deviations.

22. Independent inventory counts — coefficients -3, 1

Invented setup. Let independent X and Y represent two independent inventory counts with means 15 and 14 and standard deviations 4 and 5. A derived score is W=-3X + 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 5²=25. The expected value is E(W)=(-3)(15)+(1)(14)=-31. Because X and Y are independent, Var(W)=(-3)²(16)+(1)²(25)=169. The standard deviation is √169≈13. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Do not confuse the variance of a combination with the variance of a sample mean; averaging multiple measurements is a related but distinct construction. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 169 but the expected value becomes -59. Explain why this changed case is not a paradox. A result that uses added variance 25 with a negative sign would be confusing an operation on values with an operation on squared deviations.

23. Separate sensor outputs — coefficients 4, -1

Invented setup. Let independent X and Y represent two separate sensor outputs with means 22 and 19 and standard deviations 5 and 6. A derived score is W=4X − 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 6²=36. The expected value is E(W)=(4)(22)+(-1)(19)=69. Because X and Y are independent, Var(W)=(4)²(25)+(-1)²(36)=436. The standard deviation is √436≈20.88061. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Suppose the two measurements shared a common random source. The simple addition of variances would then need justification, not automatic use. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 436 but the expected value becomes 107. Explain why this changed case is not a paradox. A result that uses subtracted variance 36 with a negative sign would be confusing an operation on values with an operation on squared deviations.

24. Independent modelled amounts — coefficients -4, 4

Invented setup. Let independent X and Y represent two independent modelled amounts with means 6 and 5 and standard deviations 6 and 2. A derived score is W=-4X + 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 2²=4. The expected value is E(W)=(-4)(6)+(4)(5)=-4. Because X and Y are independent, Var(W)=(-4)²(36)+(4)²(4)=640. The standard deviation is √640≈25.29822. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The final standard deviation is the positive square root of the variance, not the signed square root suggested by the original subtraction symbol. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 640 but the expected value becomes -44. Explain why this changed case is not a paradox. A result that uses added variance 64 with a negative sign would be confusing an operation on values with an operation on squared deviations.

25. Independent time readings — coefficients 1, -2

Invented setup. Let independent X and Y represent two independent time readings with means 13 and 10 and standard deviations 1 and 3. A derived score is W=1X − 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 3²=9. The expected value is E(W)=(1)(13)+(-2)(10)=-7. Because X and Y are independent, Var(W)=(1)²(1)+(-2)²(9)=37. The standard deviation is √37≈6.08276. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. When the total expected value is negative, the result remains perfectly legitimate; an expectation can lie below zero even when both variance contributions are non-negative. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 37 but the expected value becomes 33. Explain why this changed case is not a paradox. A result that uses subtracted variance 36 with a negative sign would be confusing an operation on values with an operation on squared deviations.

26. Separate batch weights — coefficients 2, 2

Invented setup. Let independent X and Y represent two separate batch weights with means 20 and 15 and standard deviations 2 and 4. A derived score is W=2X + 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 4²=16. The expected value is E(W)=(2)(20)+(2)(15)=70. Because X and Y are independent, Var(W)=(2)²(4)+(2)²(16)=80. The standard deviation is √80≈8.94427. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. A coefficient of −3 affects the mean through its negative sign but affects variance through the positive factor 9. Name that distinction explicitly. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 80 but the expected value becomes 10. Explain why this changed case is not a paradox. A result that uses added variance 64 with a negative sign would be confusing an operation on values with an operation on squared deviations.

27. Two independent trial scores — coefficients -1, -4

Invented setup. Let independent X and Y represent two two independent trial scores with means 27 and 20 and standard deviations 3 and 5. A derived score is W=-1X − 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 5²=25. The expected value is E(W)=(-1)(27)+(-4)(20)=-107. Because X and Y are independent, Var(W)=(-1)²(9)+(-4)²(25)=409. The standard deviation is √409≈20.22375. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Reverse the sign of the second coefficient and predict which of the two outputs must change. The variance will be the same; the expectation generally will not. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 409 but the expected value becomes 53. Explain why this changed case is not a paradox. A result that uses subtracted variance 400 with a negative sign would be confusing an operation on values with an operation on squared deviations.

28. Independent fictional demand levels — coefficients 3, 3

Invented setup. Let independent X and Y represent two independent fictional demand levels with means 11 and 6 and standard deviations 4 and 6. A derived score is W=3X + 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 6²=36. The expected value is E(W)=(3)(11)+(3)(6)=51. Because X and Y are independent, Var(W)=(3)²(16)+(3)²(36)=468. The standard deviation is √468≈21.63331. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. If a friend has added the standard deviations rather than the variances, square the proposed answer and compare it with the computed variance. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 468 but the expected value becomes 15. Explain why this changed case is not a paradox. A result that uses added variance 324 with a negative sign would be confusing an operation on values with an operation on squared deviations.

29. Independent temperature deviations — coefficients -2, -3

Invented setup. Let independent X and Y represent two independent temperature deviations with means 18 and 11 and standard deviations 5 and 2. A derived score is W=-2X − 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 2²=4. The expected value is E(W)=(-2)(18)+(-3)(11)=-69. Because X and Y are independent, Var(W)=(-2)²(25)+(-3)²(4)=136. The standard deviation is √136≈11.6619. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. This exercise gives only the first two moments. Unless distributions are specified, do not announce that the linear combination is normal. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 136 but the expected value becomes -3. Explain why this changed case is not a paradox. A result that uses subtracted variance 36 with a negative sign would be confusing an operation on values with an operation on squared deviations.

30. Two independent processing delays — coefficients -3, 1

Invented setup. Let independent X and Y represent two two independent processing delays with means 25 and 16 and standard deviations 6 and 3. A derived score is W=-3X + 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 3²=9. The expected value is E(W)=(-3)(25)+(1)(16)=-59. Because X and Y are independent, Var(W)=(-3)²(36)+(1)²(9)=333. The standard deviation is √333≈18.24829. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The coefficients may represent unit conversions or scaled quantities; confirm that both terms have compatible units before interpreting a combined outcome. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 333 but the expected value becomes -91. Explain why this changed case is not a paradox. A result that uses added variance 9 with a negative sign would be confusing an operation on values with an operation on squared deviations.

31. Independent component measurements — coefficients 4, -1

Invented setup. Let independent X and Y represent two independent component measurements with means 9 and 21 and standard deviations 1 and 4. A derived score is W=4X − 1Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 1²=1 and 4²=16. The expected value is E(W)=(4)(9)+(-1)(21)=15. Because X and Y are independent, Var(W)=(4)²(1)+(-1)²(16)=32. The standard deviation is √32≈5.65685. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try changing the first coefficient from a to 2a. Its contribution to variance quadruples; its contribution to the mean doubles. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 32 but the expected value becomes 57. Explain why this changed case is not a paradox. A result that uses subtracted variance 16 with a negative sign would be confusing an operation on values with an operation on squared deviations.

32. Independent route variations — coefficients -4, 4

Invented setup. Let independent X and Y represent two independent route variations with means 16 and 7 and standard deviations 2 and 5. A derived score is W=-4X + 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 2²=4 and 5²=25. The expected value is E(W)=(-4)(16)+(4)(7)=-36. Because X and Y are independent, Var(W)=(-4)²(4)+(4)²(25)=464. The standard deviation is √464≈21.54066. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Try adding a fixed constant 5 to the whole expression. Its mean increases by 5 but its variance remains identical. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 464 but the expected value becomes -92. Explain why this changed case is not a paradox. A result that uses added variance 400 with a negative sign would be confusing an operation on values with an operation on squared deviations.

33. Two independent calibration errors — coefficients 1, -2

Invented setup. Let independent X and Y represent two two independent calibration errors with means 23 and 12 and standard deviations 3 and 6. A derived score is W=1X − 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 3²=9 and 6²=36. The expected value is E(W)=(1)(23)+(-2)(12)=-1. Because X and Y are independent, Var(W)=(1)²(9)+(-2)²(36)=153. The standard deviation is √153≈12.36932. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. An expected result of zero does not imply a certain result of zero. Positive variance records the remaining fluctuation. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 153 but the expected value becomes 47. Explain why this changed case is not a paradox. A result that uses subtracted variance 144 with a negative sign would be confusing an operation on values with an operation on squared deviations.

34. Independent inventory counts — coefficients 2, 2

Invented setup. Let independent X and Y represent two independent inventory counts with means 7 and 17 and standard deviations 4 and 2. A derived score is W=2X + 2Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 4²=16 and 2²=4. The expected value is E(W)=(2)(7)+(2)(17)=48. Because X and Y are independent, Var(W)=(2)²(16)+(2)²(4)=80. The standard deviation is √80≈8.94427. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Do not confuse the variance of a combination with the variance of a sample mean; averaging multiple measurements is a related but distinct construction. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 80 but the expected value becomes -20. Explain why this changed case is not a paradox. A result that uses added variance 16 with a negative sign would be confusing an operation on values with an operation on squared deviations.

35. Separate sensor outputs — coefficients -1, -4

Invented setup. Let independent X and Y represent two separate sensor outputs with means 14 and 22 and standard deviations 5 and 3. A derived score is W=-1X − 4Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 5²=25 and 3²=9. The expected value is E(W)=(-1)(14)+(-4)(22)=-102. Because X and Y are independent, Var(W)=(-1)²(25)+(-4)²(9)=169. The standard deviation is √169≈13. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. Suppose the two measurements shared a common random source. The simple addition of variances would then need justification, not automatic use. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 169 but the expected value becomes 74. Explain why this changed case is not a paradox. A result that uses subtracted variance 144 with a negative sign would be confusing an operation on values with an operation on squared deviations.

36. Independent modelled amounts — coefficients 3, 3

Invented setup. Let independent X and Y represent two independent modelled amounts with means 21 and 8 and standard deviations 6 and 4. A derived score is W=3X + 3Y. Decide what happens to the centre and to the spread. It is not enough to say that two readings were “separately reported”; in this fictional exercise their independence is an explicit assumption.

Complete calculation. The variance inputs are 6²=36 and 4²=16. The expected value is E(W)=(3)(21)+(3)(8)=87. Because X and Y are independent, Var(W)=(3)²(36)+(3)²(16)=468. The standard deviation is √468≈21.63331. Both variance contributions are non-negative even if the mean contributions partly cancel.

Interpretation and change test. The final standard deviation is the positive square root of the variance, not the signed square root suggested by the original subtraction symbol. If you change the sign of the Y coefficient while keeping its magnitude, Var(W) remains 468 but the expected value becomes 39. Explain why this changed case is not a paradox. A result that uses added variance 144 with a negative sign would be confusing an operation on values with an operation on squared deviations.

Single-variable transformations: a shift is not a scale

37. Transform X by -4X + 3

One-variable exercise. Suppose X has E(X)=8 and standard deviation 1, so its variance is 1. Set U=-4X + 3. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-4)(8)+(3)=-29. The added constant changes only the centre; the variance is Var(U)=(-4)²(1)=16, so SD(U)=4. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 8. Only the expected value changes, increasing by 5; the variance remains 16. Next halve the coefficient -4 while keeping the constant fixed: the variance becomes 4. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

38. Transform X by -3X − 8

One-variable exercise. Suppose X has E(X)=11 and standard deviation 2, so its variance is 4. Set U=-3X − 8. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-3)(11)+(-8)=-41. The added constant changes only the centre; the variance is Var(U)=(-3)²(4)=36, so SD(U)=6. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by -3. Only the expected value changes, increasing by 5; the variance remains 36. Next halve the coefficient -3 while keeping the constant fixed: the variance becomes 9. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

39. Transform X by -2X + 12

One-variable exercise. Suppose X has E(X)=14 and standard deviation 3, so its variance is 9. Set U=-2X + 12. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-2)(14)+(12)=-16. The added constant changes only the centre; the variance is Var(U)=(-2)²(9)=36, so SD(U)=6. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 17. Only the expected value changes, increasing by 5; the variance remains 36. Next halve the coefficient -2 while keeping the constant fixed: the variance becomes 9. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

40. Transform X by -1X + 4

One-variable exercise. Suppose X has E(X)=17 and standard deviation 4, so its variance is 16. Set U=-1X + 4. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-1)(17)+(4)=-13. The added constant changes only the centre; the variance is Var(U)=(-1)²(16)=16, so SD(U)=4. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 9. Only the expected value changes, increasing by 5; the variance remains 16. Next halve the coefficient -1 while keeping the constant fixed: the variance becomes 4. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

41. Transform X by 1X − 5

One-variable exercise. Suppose X has E(X)=20 and standard deviation 5, so its variance is 25. Set U=1X − 5. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(1)(20)+(-5)=15. The added constant changes only the centre; the variance is Var(U)=(1)²(25)=25, so SD(U)=5. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 0. Only the expected value changes, increasing by 5; the variance remains 25. Next halve the coefficient 1 while keeping the constant fixed: the variance becomes 6.25. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

42. Transform X by 2X + 7

One-variable exercise. Suppose X has E(X)=23 and standard deviation 6, so its variance is 36. Set U=2X + 7. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(2)(23)+(7)=53. The added constant changes only the centre; the variance is Var(U)=(2)²(36)=144, so SD(U)=12. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 12. Only the expected value changes, increasing by 5; the variance remains 144. Next halve the coefficient 2 while keeping the constant fixed: the variance becomes 36. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

43. Transform X by 3X + 0

One-variable exercise. Suppose X has E(X)=26 and standard deviation 1, so its variance is 1. Set U=3X + 0. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(3)(26)+(0)=78. The added constant changes only the centre; the variance is Var(U)=(3)²(1)=9, so SD(U)=3. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 5. Only the expected value changes, increasing by 5; the variance remains 9. Next halve the coefficient 3 while keeping the constant fixed: the variance becomes 2.25. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

44. Transform X by 4X + 10

One-variable exercise. Suppose X has E(X)=29 and standard deviation 2, so its variance is 4. Set U=4X + 10. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(4)(29)+(10)=126. The added constant changes only the centre; the variance is Var(U)=(4)²(4)=64, so SD(U)=8. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 15. Only the expected value changes, increasing by 5; the variance remains 64. Next halve the coefficient 4 while keeping the constant fixed: the variance becomes 16. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

45. Transform X by 5X − 2

One-variable exercise. Suppose X has E(X)=32 and standard deviation 3, so its variance is 9. Set U=5X − 2. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(5)(32)+(-2)=158. The added constant changes only the centre; the variance is Var(U)=(5)²(9)=225, so SD(U)=15. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 3. Only the expected value changes, increasing by 5; the variance remains 225. Next halve the coefficient 5 while keeping the constant fixed: the variance becomes 56.25. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

46. Transform X by -5X + 6

One-variable exercise. Suppose X has E(X)=35 and standard deviation 4, so its variance is 16. Set U=-5X + 6. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-5)(35)+(6)=-169. The added constant changes only the centre; the variance is Var(U)=(-5)²(16)=400, so SD(U)=20. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 11. Only the expected value changes, increasing by 5; the variance remains 400. Next halve the coefficient -5 while keeping the constant fixed: the variance becomes 100. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

47. Transform X by 2X − 11

One-variable exercise. Suppose X has E(X)=38 and standard deviation 5, so its variance is 25. Set U=2X − 11. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(2)(38)+(-11)=65. The added constant changes only the centre; the variance is Var(U)=(2)²(25)=100, so SD(U)=10. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by -6. Only the expected value changes, increasing by 5; the variance remains 100. Next halve the coefficient 2 while keeping the constant fixed: the variance becomes 25. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

48. Transform X by -2X + 9

One-variable exercise. Suppose X has E(X)=41 and standard deviation 6, so its variance is 36. Set U=-2X + 9. No independence statement is needed because this question transforms just one random quantity. First decide what a shift does, and then decide what scaling does.

Calculate and verify. E(U)=(-2)(41)+(9)=-73. The added constant changes only the centre; the variance is Var(U)=(-2)²(36)=144, so SD(U)=12. A negative coefficient reflects the random deviations, but squaring removes its sign. Check whether the mean could be negative; there is no prohibition against negative expectations in a linear model.

Transfer. Replace the added constant by 14. Only the expected value changes, increasing by 5; the variance remains 144. Next halve the coefficient -2 while keeping the constant fixed: the variance becomes 36. These separate changes are a stronger test of understanding than a memorised final number. In a real problem, confirm that the constant and scaled variable share the right measurement units before using the result in context.

Eight decision clinics where a formula may or may not apply

49. The judgement call: X−X

Decide before calculating. The same random variable is subtracted from itself. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. The result is identically zero, so its variance is zero. The independent-variables formula is inapplicable because X is not independent of itself when non-degenerate. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. Avoid using Var(X)+Var(X) for this expression; two printed X symbols do not create two independent variables. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

50. The judgement call: X+5

Decide before calculating. A fixed constant 5 is added to a random quantity. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. The mean shifts up by 5 while variance stays unchanged. A deterministic constant has no random deviation around its mean. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. A phrase like ‘add five extra minutes’ affects the centre, not the random spread, if the extra five is fixed. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

51. The judgement call: 2X

Decide before calculating. One measurement is doubled. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. Its expectation doubles, standard deviation doubles in magnitude and variance quadruples. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. The squaring occurs in variance, not in expectation; explain the distinction using deviations from the mean. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

52. The judgement call: Independent X and Y

Decide before calculating. The task explicitly says two variables are independent. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. Use the squared-coefficient variance-addition rule. You still need the individual variances and the coefficients. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. Independence justifies a formula; it does not supply the numerical inputs that the question may omit. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

53. The judgement call: Unknown relationship

Decide before calculating. The task gives means and variances of X and Y but says nothing about their relationship. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. You can calculate E(X−Y) from the means, but the variance of the difference is not generally determined by those moments alone. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. Do not silently assume independence because it would make a convenient numerical answer possible. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

54. The judgement call: Normal plus normal

Decide before calculating. Independent X and Y are each declared normal. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. A linear combination is also normal, with mean and variance from the usual independent-variable rules. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. It is the additional normal-model assumption that justifies identifying the whole distribution, not merely its first two moments. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

55. The judgement call: Two standard deviations

Decide before calculating. The question asks for SD(X−Y) rather than Var(X−Y), and the variables are independent. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. Add the variances and then take the positive square root. Do not subtract or add the two standard deviations directly. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. Variances combine in squared units, then the square root restores the original measurement unit. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

56. The judgement call: Opposite coefficients

Decide before calculating. The expression aX−bY replaces aX+bY for the same independent variables. What can you infer without inventing additional assumptions? State whether the question supplies enough information to compute a numerical mean or variance, and identify the precise rule governing the expression. If the result concerns a constant rather than a fluctuating measurement, describe that distinction in words.

Correct conclusion. The Y term’s sign changes the mean contribution, while b²Var(Y) stays the same in variance. The distinction is part of the mathematical model. Even a highly accurate numerical computation would be unsupported if it used a formula whose independence or distributional assumptions were missing.

Repair and transfer. This identity can be checked by physically reflecting the Y fluctuation about its mean. Change one detail of the statement: make the supposedly random offset a fixed number, or turn a repeated X into a newly sampled independent Y. Explain which parts of your conclusion change. The purpose is to avoid treating formal symbols as detachable from the story that defines them. Re-read the original assumption aloud after the change and check the final units.

Ten fast self-check questions

• 1. Independent Var(X)=4, Var(Y)=9: Var(X−Y)=13.
• 2. Independent E(X)=10, E(Y)=7: E(X−Y)=3.
• 3. If X has variance 9, then Var(2X)=36.
• 4. If X has variance 9, then Var(−X)=9.
• 5. If X has variance 9, then Var(X+100)=9.
• 6. If independent Var(X)=4 and Var(Y)=9, then Var(3X−2Y)=36+36=72.
• 7. If E(X)=5, then E(−2X+1)=−9.
• 8. If X is non-degenerate, Var(X−X)=0, not 2Var(X).
• 9. If independent X~N(10,4) and Y~N(7,9), then X−Y~N(3,13).
• 10. Without any dependence information, Var(X−Y) is not generally determined by Var(X) and Var(Y) alone.

An error-first revision routine

Begin with the most important words in a question: “independent”, “variance”, “standard deviation”, “total”, “difference”, “scale” and “shift”. Circle the relationship between the two random quantities. On the first day, practise simple X+Y versus X−Y questions so the mean sign and variance sign cannot be confused. On day two, introduce coefficients larger than one and demand written squares before arithmetic. On day three, add constants and identify which quantity stays fixed. Day four can mix independent normal variables with moment-only questions so the learner must distinguish a full distribution from two summary statistics. On day five, use counterexamples such as X−X and ask whether the variance formula is justified.

In a small-group tutorial, have one student own the expectation ledger, another own the variance ledger, and a third challenge hidden assumptions and units. Rotate roles with changed numbers. If each learner can defend a formula and its prerequisites, the skill will transfer to both direct probability questions and real-world modelling tasks. Merely reaching a number cannot show whether they noticed that the original problem supplied independence.

Frequently asked questions

Why do variances add when I subtract two variables?

For independent variables with finite variances, the mixed fluctuation term vanishes in expectation. A negative coefficient is squared in variance, making its contribution positive. Therefore Var(X−Y)=Var(X)+Var(Y).

Does expectation also ignore the minus sign?

No. Expectation is linear and keeps signed coefficients, so E(X−Y)=E(X)−E(Y).

Why do we square the coefficients?

Variance measures squared deviations, so scaling deviations by a multiplies their squares by a². A fixed added constant does not change deviations.

Can I add standard deviations directly?

Not as a general rule. For independent sums or differences, combine variances first and then take a square root to obtain the new standard deviation.

What if the variables are dependent?

Then the simple independent variance rule is not generally justified. You need further information about their relationship; do not assume independence from silence.

Does knowing the mean and variance imply a normal distribution?

No. Many distributions share the same mean and variance. A normal model must be given or justified separately.

Why can E(W) be zero while Var(W) is positive?

The average of the fluctuations may cancel while individual outcomes still vary around that centre. A zero mean does not mean the quantity is always zero.

Which H2 Mathematics syllabus section covers this?

The 2027 H2 Mathematics 9758 normal distribution section includes E(aX+b), Var(aX+b) and combinations E(aX+bY), Var(aX+bY) where X and Y are independent.

Continue the series and verify the source

• Why Sample Variance Uses n−1
• Top 10% and Inverse Normal Cutoffs
• Why Variances Add When Values Subtract: current guide
• Angle Between Two Planes and Normals

For the broader topic owner, see JC H2 Discrete Random Variables, Expectation and Variance and the complementary E(X²) versus [E(X)]² guide. Parents can use the Bukit Timah JC2 tuition pathway. The immutable three-student Mathematics tutoring reference illustrates a teaching format elsewhere, not the mathematical rules.

Official educational reference: SEAB Singapore–Cambridge H2 Mathematics 9758 syllabus (2027). The worked data and contexts are fictional. No example is an official examination question or a verified pupil outcome, and no assumption about actual independently operating systems should be taken from the imaginary classroom scenarios. This page is a reader-facing guide to one precise H2 decision.