Source: FRAPPY Student Samples, Principles of Statistics
Tags: combining random variables, sum of independent variables, variance addition, inverse normal, invNorm, find the mean, AP Stats, standard deviation of a sum
Difficulty: Intermediate Prerequisites: Normal distribution basics, z-score formula, understanding of mean and standard deviation. You should be comfortable with part 1 of these notes (Normal Distribution Probability and Binomial Connection) before working through this material.
Big picture: Once you know how to work with a single normal random variable, the next step is handling sums and combinations of several variables. This comes up whenever real measurements involve multiple components added together (e.g. a box that contains several items plus packaging). You also need to be able to work the z-score formula "backwards," starting from a desired probability and solving for an unknown mean or boundary value. Both skills are tested heavily on AP Statistics free-response questions and form the foundation for inference with sampling distributions.
TL;DR: When you add independent random variables, their means add and their variances add (not their standard deviations). To find an unknown mean given a target tail probability, use the inverse normal approach: look up the z-score for that percentile and solve the z-formula for the unknown.
Independent random variables
Two random variables are independent if knowing the value of one gives you no information about the value of the other. In simple terms, they do not influence each other at all.
Variance
The average of the squared deviations from the mean: σ² = E[(X – μ)²]. Think of it as a measure of spread, but in squared units. You need it because variances add nicely for independent variables, while standard deviations do not.
Variance addition rule
For independent random variables X₁, X₂, …, Xₙ: Var(X₁ + X₂ + … + Xₙ) = Var(X₁) + Var(X₂) + … + Var(Xₙ). In simple terms, you add the variances (the squared standard deviations), then take the square root at the end to get back to the standard deviation of the sum.
Inverse normal (invNorm)
The calculator function that takes a cumulative probability (area to the left) and returns the corresponding z-score or raw value. Think of it as "given the probability, what is the cutoff value?"
The scenario: Buckley Farms ships boxes containing 6 bags of chips. The empty box has a mean weight of 10 oz with SD = 0.05 oz. Each bag has mean weight 16.15 oz with SD = 0.12 oz. Find the mean and standard deviation of the total weight of a full box.
Finding the mean of the sum:
The total weight is: W = Box + Bag₁ + Bag₂ + Bag₃ + Bag₄ + Bag₅ + Bag₆
Means always add, regardless of independence: μ_W = 10 + 6(16.15) = 10 + 96.90 = 106.9 oz
Both student samples got this correct.
Finding the standard deviation of the sum:
Variances add for independent variables. You cannot simply add standard deviations.
Var(W) = (0.05)² + 6 × (0.12)² = 0.0025 + 6 × 0.0144 = 0.0025 + 0.0864 = 0.0889
SD(W) = √0.0889 ≈ 0.298 oz
Sample 1 did this correctly, working with squared terms under the radical.
Sample 2 made a critical error: they added the standard deviations directly (0.05 + 6 × 0.12 = 0.77 oz). This is wrong. Standard deviations do not add. This is one of the most heavily penalised mistakes on the AP exam for this topic.
Why variances add but standard deviations do not:
Variance measures spread in squared units. When you add independent sources of variability, the squared contributions stack. Taking the square root at the end converts back to the original units. If you add standard deviations directly, you overstate the combined spread because you are ignoring the fact that independent errors partially cancel each other out.
The scenario: Buckley Farms wants only 5% of bags to weigh less than 16 oz. The standard deviation stays at 0.12 oz. What mean weight should they target?
The correct approach (Sample 1):
We need P(X < 16) = 0.05, where X ~ N(μ, 0.12) and μ is unknown.
The z-score that cuts off the bottom 5% is z = –1.645 (from the z-table or invNorm(0.05) on a calculator).
Set up the equation: –1.645 = (16 – μ) / 0.12
Solve: –1.645 × 0.12 = 16 – μ –0.1974 = 16 – μ μ = 16 + 0.1974 μ = 16.1974 oz
Sample 1 drew a helpful diagram showing the normal curve with 5% shaded in the left tail below 16, and μ to the right of 16. This is good practice on free-response questions because it shows the grader you understand the geometry of the problem.
Sample 2's error:
Sample 2 set up 0.05 = (X – 16) / 0.12, treating 0.05 as a z-score rather than as a probability. The value 0.05 is the area under the curve, not the z-score. This gave them an answer of 16.006 oz, which is far too low to achieve the desired result. You can verify: if μ = 16.006, then z = (16 – 16.006) / 0.12 = –0.05, and P(Z < –0.05) ≈ 0.48, meaning about 48% of bags would be underweight, nowhere near the 5% target.
Mean of a sum of independent variables:
μ_(X₁ + X₂ + … + Xₙ) = μ₁ + μ₂ + … + μₙ
Variance of a sum of independent variables:
σ²_(X₁ + X₂ + … + Xₙ) = σ₁² + σ₂² + … + σₙ²
Standard deviation of the sum:
σ_sum = √(σ₁² + σ₂² + … + σₙ²)
Inverse normal setup:
Given a target probability p in the left tail, find the corresponding z from a table or invNorm(p), then solve:
z = (x – μ) / σ
for whichever quantity is unknown (x, μ, or σ).
Combining random variables is how engineers calculate tolerances for assembled products. If a device has ten components, each with its own manufacturing variability, the total variability of the assembled product comes from the variance addition rule, not from adding standard deviations. This is why the assembled product is more predictable than you might expect: independent errors partially offset.
The inverse normal technique is used in quality engineering to set process targets. If a specification says no more than 5% of items may fall below a minimum, you solve for the process mean that achieves that target given the process variability.
The single most common error: adding standard deviations instead of variances. On every AP exam that tests this topic, a large number of students make this mistake. The rule is: square first, add, then take the square root.
Students sometimes apply the variance addition rule to variables that are not independent. The rule only works for independent variables. If the variables are correlated, you need covariance terms.
On inverse normal problems, students confuse the probability (area) with the z-score. The value 0.05 is a probability. The z-score for the 5th percentile is –1.645. These are different numbers with different roles in the formula.
When solving for an unknown mean in the inverse normal setup, students sometimes get the sign of the z-score wrong. Draw a picture. If the target value (16 oz) is in the left tail, the z-score is negative because 16 is below the mean you are solving for.
⚠️ "Add variances, not standard deviations" is essentially guaranteed to appear on the AP exam. Graders specifically look for whether you squared before adding.
⚠️ On free-response questions involving inverse normal, always show: (1) the z-score you used and where it came from, (2) the equation setup, (3) the algebraic solution. A bare calculator answer with no work shown will not earn full marks.
⚠️ Drawing a diagram on inverse normal problems is strongly recommended. It demonstrates understanding and helps you avoid sign errors.
⚠️ For combining random variables, explicitly state the assumption of independence. If the problem says the variables are independent, cite that. If it does not, note that you are assuming independence.
True or false: If X has SD = 3 and Y has SD = 4, then X + Y has SD = 7.
A: False. Var(X + Y) = 9 + 16 = 25, so SD(X + Y) = 5 (assuming independence).
Fill in the blank: The z-score that corresponds to the bottom 5% of a standard normal distribution is approximately ________.
A: –1.645
True or false: When finding an unknown mean using the inverse normal method, you can plug the probability directly into the z-score formula as if it were a z-score.
A: False. You must first convert the probability into a z-score using a table or invNorm, then use that z-score in the formula.
Q: Package A has a mean weight of 5 oz (SD = 0.1 oz) and package B has a mean weight of 8 oz (SD = 0.2 oz). If the two packages are independent, what is the mean and standard deviation of their combined weight?
A: Mean = 5 + 8 = 13 oz. Variance = (0.1)² + (0.2)² = 0.01 + 0.04 = 0.05. SD = √0.05 ≈ 0.224 oz.
Q: A student adds three identical items, each with SD = 0.12, and reports the combined SD as 0.36. What should the correct answer be?
A: Var = 3 × (0.12)² = 3 × 0.0144 = 0.0432. SD = √0.0432 ≈ 0.208. The student incorrectly added standard deviations.
Q: You want P(X < 50) = 0.10 for a normally distributed variable with σ = 3. What mean should you target?
A: The z-score for the 10th percentile is –1.282. Set up: –1.282 = (50 – μ) / 3. Solve: –3.846 = 50 – μ, so μ = 53.846.
Q: Why did Sample 2 get part (d) wrong in the FRAPPY?
A: They treated the probability 0.05 as if it were a z-score and plugged it directly into the z-formula. The correct z-score for the 5th percentile is –1.645, not 0.05.
The variance addition rule reappears when you study the sampling distribution of a sample mean: the standard deviation of the sample mean is σ / √n, which comes from the same principle (the variance of a sum of n independent observations is nσ², divided by n² when you take the average, giving σ² / n).
Inverse normal reasoning is the backbone of confidence intervals. When you compute a 95% confidence interval, you are using z* = 1.96, which is the inverse normal value for 97.5% cumulative probability. The algebra is the same, just repackaged.
Related Terms / Search Tags: combining random variables, sum of independent random variables, variance addition rule, add variances not standard deviations, standard deviation of a sum, inverse normal, invNorm TI-84, find the mean given probability, AP Statistics Chapter 6, FRAPPY, Buckley Farms, quality control statistics, process mean calculation, z-score backwards