← Back to the Introduction to Statistics outline The Course Maker
← Introduction to Statistics outline
Week 10 · Lecture outline

Week 10 — Lecture Outline · Sampling Distributions

Introduction to Statistics · MATH 11 Fall 2026 · Prof. Rivera Fictional sample

Course: Introduction to Statistics (MATH 11) · Silver Oak University (fictional sample) · Prof. Rivera
Objectives covered: Objective 5 — Use normal and sampling distributions to reason about variability.
SLOs touched: A (reason quantitatively from data) · B (communicate results to a non-technical audience)
Meeting pattern: 2 sessions × 75 min = 150 min. Segment minutes below total ~150; scale to your own pattern.

Week at a Glance
The week's big question "If we'd grabbed a different sample, we'd have gotten a different average — so how much does the average itself bounce around, and when can we trust it?"
By the end of the week, students can… (1) explain sampling variability — that a statistic like x̄ changes from sample to sample; (2) describe the sampling distribution of the sample mean x̄ — its center is the population mean μ and its spread is the standard error σ/√n; (3) compute a standard error with friendly numbers and say what it means; (4) state the Central Limit Theorem — for large n, x̄ is approximately normal whatever the population shape — and use it with a z-score and a supplied table to find a probability about x̄; (5) avoid the two classic traps — confusing σ with σ/√n, and thinking the CLT is about individual values rather than averages.
Key vocabulary sampling variability, statistic vs. parameter (callback), the sampling distribution of x̄, the mean of x̄ (= μ), the standard deviation of x̄ = the standard error (SE) = σ/√n, the Central Limit Theorem (CLT), "large n" (a working rule of thumb: n ≥ 30), z-score for a sample mean, cumulative area / probability
Materials slides (Deck 10), the week's readings + video links, a spreadsheet (Google Sheets or Excel), one approved chatbot (Gemini / Claude / ChatGPT) for the AI-critique moment and the tutorial, and the small z-table embedded in this outline (we hand students the values — they never recall them)
Timing note 8 segments, ~150 min total. Session 1 = Segments 1–4 (~75). Session 2 = Segments 5–8 (~75).

The z-table we use all week (cumulative area to the LEFT of z). Identical to the Week 9 table. Every worked example, practice item, quiz item, and assignment problem is engineered to land exactly on one of these friendly values — and we always pre-compute the standard error σ/√n with friendly numbers so the z does too. We supply these numbers; we never ask a student — or a chatbot — to recall or estimate them.

z −3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0 +0.5 +1.0 +1.5 +2.0 +2.5 +3.0
area to LEFT .0013 .0062 .0228 .0668 .1587 .3085 .5000 .6915 .8413 .9332 .9772 .9938 .9987
area to RIGHT .9987 .9938 .9772 .9332 .8413 .6915 .5000 .3085 .1587 .0668 .0228 .0062 .0013

The three moves are unchanged: area to the left = read the table; area to the right = 1 − (area to the left); area between two z's = (left of the bigger) − (left of the smaller). Left + right always = 1. The one new step this week: before standardizing a sample mean, divide σ by √n to get the right spread.

Segment 1 — Hook & the Promise (8 min) · Session 1 opens

Hook. Run a tiny live poll. "Everyone guess the average number of hours this class slept last night — but I'm only going to ask the first four people in row one." Get four numbers, average them on the board (say it lands on 6.4). Then: "New sample — first four in row three." Average those (say 7.1). Same class, same true average out there somewhere — two different answers.
- "Nobody made a mistake. The average moved because the sample moved. That wobble has a name — sampling variability — and this week we stop being surprised by it and start measuring it."
- "Here's the twist that runs the whole second half of this course: that wobble is predictable. The averages don't bounce around wildly — they cluster, in a bell, around the truth. We can say exactly how tightly."

The promise (write it on the board): "By the end of this week you can take a population's mean and standard deviation, a sample size n, and tell me how a sample average x̄ behaves — where it centers, how much it spreads, and the chance it lands in any range — using nothing but σ/√n, a z-score, and a small table."

Why it matters line (memory hook): "One sample gives you a number. Sampling distributions tell you how much to trust that number — and a bigger sample buys you a tighter answer."

Segment 2 — Sampling Variability: a Statistic Is a Moving Target (18 min)

Plain language first.
- Back in Week 1 we split parameter (a fixed number describing the whole population — like the true mean μ) from statistic (a number from your sample — like the sample mean x̄). Here's the part we glossed over then: the statistic isn't one number — it's a different number every time you draw a fresh sample. That moving-around is sampling variability.
- So x̄ is itself a random variable (Week 6 idea): take a sample, get an x̄; take another, get another x̄. If you did this thousands of times and made a histogram of all those x̄ values, you'd get a brand-new distribution — the sampling distribution of the sample mean. It's not a distribution of people; it's a distribution of averages.
- Notation note (after the idea): the population has mean μ and standard deviation σ. The sampling distribution of x̄ has its own center and its own spread, and the next two segments are entirely about what those are.

Memory hook (put it on a slide):

One population, many samples, many x̄'s. The sampling distribution is the histogram of all those averages.

One fully worked example (do it out loud).

Setup: Imagine a campus where the true average daily coffee spend is μ = $4.00. You don't know that; you take a sample of 25 students and get x̄ = $4.30. Your friend takes a different 25 and gets x̄ = $3.80. A third group gets $4.05.
- None of these is "wrong." Each is an honest statistic from an honest sample — they differ because of sampling variability alone.
- Pile up thousands of such x̄'s and the histogram would be centered near $4.00 (the truth) and far narrower than the spread of individual students' spending. Averages are calmer than individuals — and Segments 3–4 say exactly how calm.

Land the key idea: a statistic is a moving target; the pattern of its movement (its sampling distribution) is what lets us reason about how close any one sample got to the truth.

Segment 3 — The Sampling Distribution of x̄: Center μ, Spread σ/√n (28 min)

Plain language first — two facts pin it down. Just like a normal curve needed a center and a spread, the sampling distribution of x̄ needs two numbers, and both come straight from the population:

  • Center (the mean of x̄) = μ. The averages pile up around the true population mean. On average, x̄ neither overshoots nor undershoots μ — it's "aimed right." (The fancy word is unbiased; the plain idea is "centered on the truth.")
  • Spread (the standard deviation of x̄) = the STANDARD ERROR = σ/√n. This is the whole engine of the week. The averages spread less than individuals do — and the more people you average over (bigger n), the tighter they cluster, because √n is in the denominator.

Memory hook: "Center stays put at μ; the spread shrinks to σ over root-n." The standard error is just "the standard deviation of the average."

Why √n and not n (one sentence, plain): doubling your sample doesn't halve the wobble — you have to quadruple n to halve the standard error, because it's √n that does the work. "To cut the wobble in half, take four times as many people."

One fully worked example (pre-computed, friendly numbers).

Setup: Individual students' daily coffee spend has population mean μ = $4.00 and standard deviation σ = $2.00. We take samples of size n = 100.
- Center of the sampling distribution of x̄ = μ = $4.00.
- Standard error = σ/√n = 2.00 / √100 = 2.00 / 10 = $0.20.
- So sample averages of 100 students bounce around $4.00 with a spread of only 20 cents — even though individuals swing by $2.00. The averaging shrank the spread by a factor of √100 = 10. "Ten times more people, ten times less wobble."

A second worked example (different friendly numbers, to drill σ/√n):

Battery lifetimes have μ = 500 hours, σ = 60 hours. For samples of n = 9 batteries: SE = 60 / √9 = 60 / 3 = 20 hours. For samples of n = 36: SE = 60 / √36 = 60 / 6 = 10 hours. Quadrupling n (9 → 36) cut the standard error in half (20 → 10). Make them watch the √n do that.

The habit to give students: every sampling-distribution problem starts the same way — write down μ (the center) and compute σ/√n (the spread) before anything else. Those two numbers are the whole description.

Segment 4 — Misconceptions + Quick Interaction (21 min) · Session 1 closes (~75)

Name the misconceptions out loud, then cure each:

  • ❌ "The spread of the averages is just σ." (The single biggest error of the week.)*
    ✅     Cure: σ is the spread of individuals; the spread of the average is σ/√n, which is smaller. Mixing them up means using a number that's too big and getting every probability wrong. "Individuals spread by σ; averages spread by σ over root-n. Always divide by √n for a mean." Put both side by side on the board and circle the √n.
  • ❌ "The Central Limit Theorem says the data become normal."
    ✅ Cure: the CLT is about the sampling distribution of x̄ (the averages), not the raw data. A skewed population stays skewed no matter how big n is — it's the distribution of the sample mean that turns normal. "The CLT bells the averages, not the individuals." (We'll hammer this in Segment 6.)
  • ❌ "A bigger sample makes the population's standard deviation smaller."
    ✅ Cure: σ is a fixed fact about the population — sample size can't touch it. What shrinks with bigger n is the standard error σ/√n, the spread of the averages. "n moves the SE, never σ."
  • ❌ "x̄ will equal μ if my sample is big enough."
    ✅ Cure: x̄ is still random — a bigger n makes it land closer to μ on average (smaller SE), but it won't sit exactly on μ. "Bigger n tightens the cluster; it doesn't freeze the value."

Interaction — Think-Pair-Share (standard-error reflexes, ~12 min):
Put μ = 500, σ = 60 on a slide. Students answer solo (30 sec), compare with a neighbor (1 min), class votes / calls out:
1. SE for n = 4? (60 / √4 = 60 / 2 = 30)
2. SE for n = 9? (60 / 3 = 20)
3. SE for n = 36? (60 / 6 = 10)
4. To get the SE down to 6, what n do you need? (60 / √n = 6 → √n = 10 → n = 100)
Debrief #4, which always stretches the room: solve σ/√n for n. They should see that halving the SE again and again costs ever-larger jumps in n — the √n tax. "Precision is expensive; each extra decimal of certainty costs four times the sample."

Segment 5 — The Central Limit Theorem (25 min) · Session 2 opens

Hook back in: "Friday we found the center and spread of the averages. But to turn that into a probability — 'what's the chance the average lands above ___?' — we need to know the shape. That's the miracle the Central Limit Theorem hands us."

Plain language first — what the CLT says.
- The Central Limit Theorem (CLT): when the sample size n is large, the sampling distribution of x̄ is approximately normal — no matter what shape the population has. Skewed, lumpy, weird — average enough of it together and the averages form a bell.
- So once n is large, x̄ has all three pieces nailed down: it's normal, centered at μ, with spread σ/√n. That's exactly a normal distribution N(μ, σ/√n) — and everything we learned in Week 9 (z-scores, the table, left/right/between) now works on sample means, with one change: divide by σ/√n instead of σ.
- "How large is large?" A common working rule of thumb: n ≥ 30 is plenty for the CLT to kick in for most populations (more if the population is wildly skewed; less if it's already bell-ish). Treat 30 as a guideline, not a magic line.

Memory hook: "Average enough of anything and you get a bell — centered at μ, spread σ/√n. The CLT is the doorway to inference."

Why this is the big deal (one line): it means we can compute probabilities about a sample average even when we have no idea what the population's shape is. Most of the rest of the course leans on this one theorem.

One fully worked example (CLT probability, pre-computed to land on the table).

Setup: Individual purchase amounts at a campus café are right-skewed (a few big orders) with μ = $4.00, σ = $2.00. A manager takes a random sample of n = 100 purchases. What is the probability the sample mean x̄ is less than $3.60?
1. Shape: n = 100 ≥ 30, so by the CLT, x̄ is approximately normal — even though the raw purchases are skewed. (Name it out loud: "we're allowed to use the bell here only because of the CLT.")
2. Center & spread: mean of x̄ = μ = $4.00; standard error = σ/√n = 2.00 / √100 = 2.00 / 10 = $0.20.
3. Standardize the sample mean: z = (x̄ − μ) ÷ (σ/√n) = (3.60 − 4.00) ÷ 0.20 = (−0.40) ÷ 0.20 = −2.0.
4. Read the table: area to the LEFT of z = −2.0 is .0228. So P(x̄ < $3.60) ≈ .0228, about a 2.28% chance. "Even though one person's spend below $3.60 is common, the average of 100 landing that low is rare — averaging squeezes the spread."

A second worked example — a "between" probability (same setup, drill the moves):

What's the probability x̄ lands between $3.80 and $4.40?
- 3.80 → z = (3.80 − 4.00) ÷ 0.20 = −1.0; 4.40 → z = (4.40 − 4.00) ÷ 0.20 = +2.0.
- Between = (left of +2.0) − (left of −1.0) = .9772 − .1587 = .8185 → about 81.85%. Most samples of 100 average within that band.

The interpretation drill to give students: every CLT probability is a three-line ritual — (1) "CLT says x̄ is normal," (2) write μ and σ/√n, (3) standardize with σ/√n and read the table. Say the answer in a sentence: "there's about a 2% chance the average of 100 purchases is below $3.60."

Segment 6 — CLT vs. Individuals; "Average enough of anything" (16 min)

Plain language first — the distinction that separates an A from a C this week.
- Ask the class to hold two pictures at once: the population of individual values (could be skewed, bimodal, anything) and the sampling distribution of x̄ (a tidy bell once n is large). They are different distributions with different spreads.
- A z-score about one individual uses σ (Week 9). A z-score about a sample average uses σ/√n (this week). Same z-table, different denominator. "One value? Divide by σ. An average? Divide by σ/√n."

One fully worked contrast (make them feel the gap):

Daily steps for individuals have μ = 8,000, σ = 3,000 and are right-skewed.
- One person: the chance a single person walks under 6,500 steps uses σ = 3,000: z = (6500 − 8000) ÷ 3000 = −0.5 → left area .3085 ≈ 30.85% — and strictly we'd hesitate because individuals are skewed, not normal (Week 9's "is it even a bell?" check).
- A sample of n = 36: the chance the average of 36 people is under 7,500 uses σ/√n = 3000 / √36 = 3000 / 6 = 500: z = (7500 − 8000) ÷ 500 = −1.0 → left area .1587 ≈ 15.87% — and here the CLT earns us the bell even though individuals are skewed. Two different denominators, two different questions, two different answers.

Misconception + cure (re-up the headline trap):
- ❌ "It's skewed, so I can't use a normal curve at all."
✅ Cure: for an individual, right — be careful (Week 9). For a sample average with large n, the CLT makes x̄ normal despite the skew. The shape of the averages is what changed.

Quick mini-check (genuinely clarifying, ~3 min): "A population is heavily right-skewed. (a) Is the histogram of individuals normal? (b) Is the histogram of sample means for n = 50 approximately normal?" Answers: (a) no — skew stays; (b) yes — the CLT. Surface that the only thing that changed is "individuals vs. averages."

Segment 7 — Technology Workflow & AI-Critique (16 min)

Plain language — let the spreadsheet do the table, but know what you fed it.
- The single most common machine mistake on these problems is feeding NORM.DIST the population σ when the question is about an average. The fix is to compute the standard error first and use that as the "SD."

Technology workflow — a CLT probability in a spreadsheet (exact steps, using μ = 4, σ = 2, n = 100):
1. Standard error first: in a cell, =2/SQRT(100) → 0.2. (Always compute σ/√n before anything else.)
2. Probability the average is below a value (left area): =NORM.DIST(3.60, 4, 0.2, TRUE) → 0.0228 (matches the table at z = −2). Note the third argument is the standard error 0.2, not σ = 2.
3. Probability the average is above a value (right area): =1 - NORM.DIST(4.40, 4, 0.2, TRUE) → 0.0228 (z = +2).
4. Standardize a sample mean directly: =STANDARDIZE(3.60, 4, 0.2) → −2 (that's the z-score for x̄). Google Sheets and Excel are identical here.

AI-critique moment (students verify, not consume):

Paste this to an approved chatbot: "Purchase amounts have mean 4 and standard deviation 2. For a sample of 100, what's the probability the sample mean is below 3.60?"
Then check it against the table. The correct answer is .0228 (SE = 2/√100 = 0.2; z = −2). Chatbots frequently (a) forget to divide by √n and use σ = 2, getting z = (3.60 − 4)/2 = −0.2 and a badly wrong probability; (b) answer about an individual instead of the average; or (c) flip left and right. Your job all semester: the tool drafts, you judge — and the giveaway here is whether it used the standard error or the raw σ. This is exactly how the weekly Lecture Tutorial and the adaptive Assignment work — you'll catch the model, not trust it.

Segment 8 — Callback, Hand-off & Tease (10 min) · Session 2 closes (~75)

Callback (tie the week together in three lines):
- Sampling variability → a statistic like x̄ is a moving target; its pattern of movement is the sampling distribution.
- Center & spread → the averages center at μ and spread by the standard error σ/√n — tighter than individuals, and tighter still as n grows.
- The CLT → for large n, x̄ is approximately normal whatever the population's shape, so a z-score (using σ/√n) and the table give the probability of any range. "One sample gives a number; the sampling distribution tells you how much to trust it."

Tease next week: "We now know how a sample average behaves if we know μ and σ. But in real life we don't know them — that's the whole point of taking a sample! Next week we flip the machine around: use a single sample to build a confidence interval — a plain-English 'we're 95% sure the true mean is in here.' The standard error you mastered today is the beating heart of it."

Hand-off (the week's graded work):
- Lecture Tutorial 10 (AI tutor, share-link submission) — sampling variability, the center/spread of x̄, the standard error, and the CLT (with the supplied table).
- Practice 10 (ungraded reps) — floor-difficulty standard-error and CLT-probability drills.
- Quiz 10 (end of week) — 10 items across sampling variability, SE = σ/√n, the CLT, and the sampling distribution of x̄.
- Discussion 10 ("How much would this average move?" — find a real reported sample average and reason about its variability) and Assignment 10 (four coached problems; submit the AI's self-scored report + chat link).

Instructor FAQ — Common Stumbles

Student says / does Quick cure
Uses σ as the spread of the average. The spread of an average is the standard error σ/√n, not σ. σ is for individuals. Write both, circle the √n: "for a mean, always divide by root-n."
"The CLT makes the data normal." No — it makes the sampling distribution of x̄ (the averages) normal. A skewed population stays skewed; only the averages form a bell.
Thinks bigger n shrinks σ. σ is a fixed property of the population; n can't change it. Bigger n shrinks the standard error σ/√n — the spread of the averages.
Forgets √ and uses σ/n. It's σ over √n. With σ = 60, n = 9: 60/√9 = 60/3 = 20 (not 60/9). The square root is the whole point — quadruple n to halve the SE.
Standardizes a sample mean with σ. For x̄, z = (x̄ − μ) ÷ (σ/√n). Compute the standard error first, then divide. Using σ here is the week's signature mistake.
"x̄ will equal μ with a big enough sample." x̄ is still random. Bigger n shrinks the SE so x̄ lands closer to μ on average — it doesn't pin x̄ exactly on μ.
Reports the left area when asked "above." "Above" = right area = 1 − (left). Same rule as Week 9; draw the curve and shade the asked-for side.
Applies the CLT to a tiny n on a wild population. The CLT needs large n (rule of thumb n ≥ 30; more for heavy skew). For small n from a non-normal population, the bell isn't guaranteed — say so.
Trusts a chatbot's probability without checking the SE. The giveaway is the denominator: did it use σ/√n or raw σ? Recompute the standard error by hand and check the z against the supplied table.

Scope flag

This outline stays within Objective 5. The "n ≥ 30" rule of thumb, the unbiased aside, the STANDARDIZE/NORM.DIST-with-SE spreadsheet functions, and the individual-vs-average contrast in Segment 6 are added context (not strictly required by the objective) — kept because they prevent the predictable σ-vs-σ/√n disaster and set up confidence intervals in Week 11. Trim them for a leaner 60-minute version; the standard error σ/√n and the CLT (x̄ is normal, centered μ, spread σ/√n) are the non-negotiable core.

~ Prof. Rivera's edition · Fall 2026 · built with thecoursemaker.com