Why do variances add when I subtract two random variables?

Subtracting still combines two independent sources of randomness, and combining uncertainty always increases total variability. The negative sign affects the mean but gets squared away in the variance calculation, since (-1)^2 = 1. Adding variances for both sums and differences is the correct rule.

Can I just add standard deviations instead of variances?

No. Standard deviations do not add. You must square each to get variance, add the variances, then take the square root. Adding standard deviations directly is the most common mistake on this topic and produces a wrong answer.

What is the difference between 2X and X_1 + X_2?

2X scales a single variable, giving variance 2^2^2 = 4^2. X_1 + X_2 adds two independent copies, giving variance ^2 + ^2 = 2^2. They have the same mean but different spreads, so read the problem carefully.

Do I always need independence to combine random variables?

Means always add or subtract without any independence assumption. But the simple rule of adding variances requires independence. If the variables are not independent, you cannot use _X^2 + _Y^2, and the AP exam will not ask you to combine variances without stating independence.

AP-STATS-4.8

U4.8 Combining Random Variables

Master AP Statistics 4.8: transform random variables with Y=a+bX and combine independent variables where means add or subtract but variances always add.

What you'll do in this lesson

A voice-first session with the Crimsora tutor on U4.8 Combining Random Variables, then targeted practice and FRQs — with the tutor adapting to where you get stuck.

Start U4.8 Combining Random Variables free

What this lesson covers

By now you know how to find the mean and standard deviation of a single random variable. Topic 4.8 asks a bigger question: what happens when you scale a variable, shift it, or add two variables together? These situations show up constantly — combining the weights of two packages, scaling a temperature from Celsius to Fahrenheit, or finding the difference between two test scores.

The rules are short but easy to misapply. The single most tested idea in this lesson is that when you combine independent random variables, you add their variances — even when you subtract their means. Get comfortable moving between standard deviation and variance, and this topic becomes reliable points on the exam.

Transforming a Single Random Variable

A linear transformation takes a random variable

X

and produces a new variable

Y = a + bX

, where

a

shifts the distribution and

b

scales it. The rules follow directly from how center and spread respond to these operations.

Adding a constant

a

slides every value over, so it changes the mean by

a

but does not change the spread. Multiplying by

b

stretches or compresses the distribution, which affects both center and spread.

Quantity	Rule	Why
Mean	$\mu_Y = a + b\mu_X$	shifting and scaling both move the center
Standard deviation	$\sigma_Y = \|b\|\sigma_X$	only scaling changes spread; sign ignored
Variance	$\sigma_Y^2 = b^2\sigma_X^2$	variance scales by the square

Two things trip students up. First, adding a constant never changes standard deviation or variance — spread is about distances between values, and shifting everything equally leaves those distances unchanged. Second, the absolute value

|b|

matters because standard deviation can never be negative. If you multiply by

-2

, the spread grows by a factor of

2

, not

-2

. On the exam, you may be given

\mu_X

and

\sigma_X

and asked for the mean and SD after a real-world rescaling, such as converting units or applying a fee.

Combining Two Random Variables: Means

When you form a sum

S = X + Y

or a difference

D = X - Y

, the means always combine in the obvious way, and this is true whether or not

X

and

Y

are independent.

\mu_{X+Y} = \mu_X + \mu_Y \qquad \mu_{X-Y} = \mu_X - \mu_Y

More generally, for a linear combination

T = aX + bY

, the mean is

\mu_T = a\mu_X + b\mu_Y

. This property is called linearity of expectation, and it holds unconditionally.

A common exam setup gives you the mean number of minutes for two separate tasks and asks for the expected total time. You simply add the means. If a problem describes the difference between two quantities — say, how much taller one plant is than another on average — subtract the means. The subtlety is never in the means; it is always in the spread, which the next section handles. Keep the mean calculation clean and simple so you can focus your attention on the variance rule, where mistakes actually happen.

Combining Two Random Variables: Variances Always Add

Here is the rule that defines this topic: when

X

and

Y

are independent, the variances add for both sums and differences.

\sigma_{X+Y}^2 = \sigma_X^2 + \sigma_Y^2 \qquad \sigma_{X-Y}^2 = \sigma_X^2 + \sigma_Y^2

Notice both formulas are identical. Even when you subtract the variables, you add the variances. The reason is that subtracting introduces more uncertainty, not less — combining two sources of random variation always increases total variability, regardless of the sign.

To get the standard deviation, take the square root at the very end:

\sigma_{X-Y} = \sqrt{\sigma_X^2 + \sigma_Y^2}

The number one error students make is adding standard deviations directly, like

\sigma_{X+Y} = \sigma_X + \sigma_Y

. This is wrong. You must convert to variance, add, then convert back. A second error is trying to subtract variances for a difference; this can produce a negative number under the square root, a clear signal you broke the rule.

Independence is required. If the exam does not state or imply that the variables are independent, you cannot add the variances using this simple rule. For a general linear combination

T = aX + bY

with independence,

\sigma_T^2 = a^2\sigma_X^2 + b^2\sigma_Y^2

How the Exam Tests This

AP questions on 4.8 usually appear as short multiple-choice items or as a part of a larger free-response problem. Watch for these signature setups.

One classic form combines a transformation with a combination: you scale each variable first, then add or subtract. Apply

\sigma_T^2 = a^2\sigma_X^2 + b^2\sigma_Y^2

, remembering to square the coefficients.

Situation	Mean	Variance (independent)
$Y = a + bX$	$a + b\mu_X$	$b^2\sigma_X^2$
$X + Y$	$\mu_X + \mu_Y$	$\sigma_X^2 + \sigma_Y^2$
$X - Y$	$\mu_X - \mu_Y$	$\sigma_X^2 + \sigma_Y^2$
$aX + bY$	$a\mu_X + b\mu_Y$	$a^2\sigma_X^2 + b^2\sigma_Y^2$

A frequent trap distinguishes between the sum of two independent copies of a variable,

X_1 + X_2

, and a single scaled variable,

2X

. These are different. For

X_1 + X_2

, variance is

\sigma^2 + \sigma^2 = 2\sigma^2

. For

2X

, variance is

2^2\sigma^2 = 4\sigma^2

. Read carefully: two separate objects means add, one object scaled means use

b^2

. On free-response, always justify independence before adding variances to earn full credit.

Key terms

Linear transformation.: An operation of the form $Y = a + bX$ that shifts a random variable by $a$ and scales it by $b$ .
Linearity of expectation.: The property that $\mu_{aX+bY} = a\mu_X + b\mu_Y$ , which holds whether or not the variables are independent.
Independent random variables.: Two variables where the value of one gives no information about the other; required for the simple variance-addition rule.
Variance.: The square of the standard deviation, $\sigma^2$ ; the quantity that adds when combining independent random variables.
Standard deviation.: A measure of spread, always non-negative; equals the square root of variance and must be found last when combining variables.
Linear combination.: An expression $aX + bY$ built from scaled random variables, with variance $a^2\sigma_X^2 + b^2\sigma_Y^2$ under independence.

Worked example

A coffee shop's morning revenue

X

has mean

\mu_X = 400

dollars and standard deviation

\sigma_X = 30

dollars. Its afternoon revenue

Y

has mean

\mu_Y = 250

dollars and standard deviation

\sigma_Y = 40

dollars. Morning and afternoon revenues are independent. Find the mean and standard deviation of the total daily revenue

T = X + Y

, and of the difference

D = X - Y

Start with the means, which add for a sum and subtract for a difference.

For the total:

\mu_T = \mu_X + \mu_Y = 400 + 250 = 650

dollars.

For the difference:

\mu_D = \mu_X - \mu_Y = 400 - 250 = 150

dollars.

Now the spread. Because the two revenues are independent, variances add for both the sum and the difference. First convert standard deviations to variances:

\sigma_X^2 = 30^2 = 900

and

\sigma_Y^2 = 40^2 = 1600

.

For the total:

\sigma_T^2 = 900 + 1600 = 2500

, so

\sigma_T = \sqrt{2500} = 50

dollars.

For the difference:

\sigma_D^2 = 900 + 1600 = 2500

as well, so

\sigma_D = \sqrt{2500} = 50

dollars.

Notice both the sum and the difference have the same standard deviation of 50 dollars, even though the means differ. That is the key lesson: subtracting the variables does not subtract the variability. A tempting wrong answer adds standard deviations directly to get

30 + 40 = 70

, which ignores the required conversion to variance and back.

Practice questions

Random variables

X

and

Y

are independent with

\sigma_X = 5

and

\sigma_Y = 12

. What is the standard deviation of

X - Y

$7$
$13$
$17$
$-7$

Answer: $13$

Variances add even for a difference:

\sigma_{X-Y}^2 = 5^2 + 12^2 = 25 + 144 = 169

. The standard deviation is

\sqrt{169} = 13

. Choosing 7 comes from subtracting standard deviations, and 17 comes from adding them; both skip the required variance step.

The weight of one apple has mean

0.3

pounds and standard deviation

0.05

pounds. A bag holds 4 apples whose weights are independent. Find the mean and standard deviation of the total weight of the bag.

Answer: Mean $= 1.2$ pounds; standard deviation $\approx 0.1$ pounds.

The total is

X_1 + X_2 + X_3 + X_4

, four independent copies. The mean is

4 \times 0.3 = 1.2

pounds. Variances add:

\sigma^2 = 4 \times 0.05^2 = 4 \times 0.0025 = 0.01

, so

\sigma = \sqrt{0.01} = 0.1

pounds. This differs from

4X

(one apple scaled), which would give

\sigma = 4 \times 0.05 = 0.2

pounds — a common trap.

A random variable

X

has

\mu_X = 20

and

\sigma_X = 4

. Define

Y = 100 - 3X

. Find

\mu_Y

and

\sigma_Y

Answer: $\mu_Y = 40$ and $\sigma_Y = 12$ .

Using

\mu_Y = a + b\mu_X

with

a = 100

and

b = -3

\mu_Y = 100 + (-3)(20) = 100 - 60 = 40

. For spread,

\sigma_Y = |b|\sigma_X = |-3| \times 4 = 12

. The absolute value ensures the standard deviation stays positive even though

b

is negative; the constant 100 has no effect on spread.

FAQ

Why do variances add when I subtract two random variables?: Subtracting still combines two independent sources of randomness, and combining uncertainty always increases total variability. The negative sign affects the mean but gets squared away in the variance calculation, since $(-1)^2 = 1$ . Adding variances for both sums and differences is the correct rule.
Can I just add standard deviations instead of variances?: No. Standard deviations do not add. You must square each to get variance, add the variances, then take the square root. Adding standard deviations directly is the most common mistake on this topic and produces a wrong answer.
What is the difference between $2X$ and $X_1 + X_2$ ?: $2X$ scales a single variable, giving variance $2^2\sigma^2 = 4\sigma^2$ . $X_1 + X_2$ adds two independent copies, giving variance $\sigma^2 + \sigma^2 = 2\sigma^2$ . They have the same mean but different spreads, so read the problem carefully.
Do I always need independence to combine random variables?: Means always add or subtract without any independence assumption. But the simple rule of adding variances requires independence. If the variables are not independent, you cannot use $\sigma_X^2 + \sigma_Y^2$ , and the AP exam will not ask you to combine variances without stating independence.

Learn this with a teacher, not a page

The Crimsora tutor teaches U4.8 Combining Random Variables live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.

Try a live session free