Join · scan

OTTO-INF1

ottolearn.org/join

Module 4 · Statistical Inference · INF-1

Sampling Distributions & the CLT
"What would a different 300 have said?"

A letter lands on Maya's desk: "You surveyed 300 of 48,000. A different 300 would've given a different average. Your 6.8 is one roll of the dice — why should anyone trust it?" The reader is right that a different 300 gives a different average — and today we meet the mathematics of exactly how different: the distribution of the averages themselves. It's the question Maya has carried since her very first survey, finally becoming answerable.

Live session · join with code OTTO-INF1 · ottolearn.org/join

Same rules as always: vote before I explain, every poll is anonymous, and changing your mind after discussion is the goal.

Today · one letter, five answers

Taking the skeptic seriously — with mathematics.

Meet the distribution of averages, find its center and its spread, learn the theorem that fixes its shape, see what sample size buys — then compute how rare any average is. Every step starts with your vote.

Recall · last class · 45 seconds

Warm-up · commit first · 60 seconds · anonymous

Hold your answer — it is today's lesson. By the end of class you'll compute exactly how far a different 300 could wander, and the number is smaller than almost everyone's first instinct.

Core concept 1 · C1

Three levels: the population, one sample, and the pile of averages

The skeptic's objection lives one level above anything we've drawn before. Keep three objects apart: the population (all values), one sample (the values Maya got), and — new today — the sampling distribution: the distribution of the averages from every possible sample.

Level 1 · The population

All 48,000 students' sleep hours. Its histogram is fixed but never fully seen; its mean \(\mu\) and standard deviation \(\sigma\) are the unknown truth Maya is after.

\(N = 48{,}000\ \cdot\ \mu,\ \sigma\)

Level 2 · One sample

Maya's 300 responses — the data she actually holds, with \(\bar{x} = 6.8\) h. A different random 300 would be a different level-2 object with a different \(\bar{x}\).

\(n = 300\ \cdot\ \bar{x} = 6.8\)

Level 3 · The sampling distribution

Imagine every possible sample of 300, each boiled down to its average — then the histogram of all those \(\bar{x}\)'s. Never directly observed; it's the theoretical object that says how far any one \(\bar{x}\) can sit from \(\mu\).

\(\mu_{\bar{x}},\ \sigma_{\bar{x}}\)

The skeptic is asking a level-3 question: "how much do the \(\bar{x}\)'s scatter?" — and level 3 is precisely the thing built to answer it. Watch all three levels side by side:

See it · C1

Three distributions, one screen

Predict first: I'll click Draw New Sample five times. Which panels change and which stay fixed? Then: the population is right-skewed — will the panel of averages be skewed too?

Population:
① Population Distribution Fixed — theoretical

All possible individual values. This is what you're sampling from.

Histogram showing the shape of the population. This panel never changes.
μ = 50, σ = 15

② Your Sample (n = 30) Changes each draw

One specific dataset you collected. Not the population — just a snapshot of it.

Histogram of n observations from one specific sample. Redraws each time you draw a new sample.
x̄ = —

③ Sampling Distribution of x̄ Fixed — theoretical

Not your data. The theoretical distribution of all possible x̄ values — if the study were repeated infinitely. You never observe this directly.

Smooth normal curve showing the theoretical sampling distribution by the CLT. This panel never changes.
Theoretical: x̄ ~ N(50, SE²)
Figure: Three distinct objects. Click ↺ Draw New Sample repeatedly — Panel ② changes every time while Panels ① and ③ remain fixed. The sampling distribution (Panel ③) is a theoretical construct, not data. Notice how much narrower it is than the population.

Quick check · the name trap

Commit first · today's summit

Core concept 2 · C2

The two dials of the pile of averages

The level-3 object has a center and a spread, and both come out of formulas — no simulation needed. These two facts hold for any population, any shape, any \(n\).

Center: averages aim true

The pile of \(\bar{x}\)'s centers exactly on the population mean: \(\mu_{\bar{x}} = \mu\). Sample means don't systematically over- or under-shoot — \(\bar{x}\) is an unbiased estimator. Maya's method aims at the truth.

\(\mu_{\bar{x}} = \mu\)

Spread: the standard error

The typical distance between a sample average and \(\mu\) is the standard error \(\text{SE} = \sigma/\sqrt{n}\). For Maya: \(\text{SE} = 1.3/\sqrt{300}\) \(\approx 0.075\) h — averages of 300 wander by minutes, not hours.

\(\sigma_{\bar{x}} = \dfrac{\sigma}{\sqrt{n}}\)

\(\sigma\) is for students, SE is for averages

One student wanders \(\pm 1.3\) h; an average of 300 wanders \(\pm 0.075\) h. Same survey, two different yardsticks — and using \(\sigma\) where the SE belongs is the most common error of the whole topic.

\(\sigma \neq \sigma_{\bar{x}}\)

The warm-up, resolved: a different 300 would almost surely land within about \(\pm 0.15\) h (two SEs) of the truth. That is why one careful survey is printable. Watch the two yardsticks live:

See it · C2

One curve for students, a much narrower one for averages

Predict first: as I drag \(n\) from 9 to 100, the individuals' curve and the averages' curve — which moves, which stays? And does the averages' curve ever get wider?

Individual observation — spread = σ = 15 Sample mean x̄ — spread = SE = σ/√n = 15.00
Two overlapping bell curves centered at μ = 50. The blue curve shows the spread of individual observations (σ = 15, fixed). The orange curve shows the spread of sample means (SE = σ/√n), which shrinks as n increases. Bracket annotations below the axis show ±σ and ±SE.
At n = 1, SE = σ — both curves are identical.
SE = 15/√1 = 15.00
Figure: The blue curve (σ = 15) never changes — it describes individual observations. The orange curve narrows as n increases, showing how sample means cluster more tightly. Drag the slider to n = 100 to see the full contrast. To halve the SE, you must quadruple n.

Practice · the yardstick for an average · fresh numbers on demand

Core concept 3 · C3

The Central Limit Theorem: the shape comes free

Center and spread are settled. The last question about the pile of averages is its shape — because only a known shape lets us read probabilities off a table. The answer is one of the most remarkable theorems in statistics.

What the theorem says

Take big enough random samples and the averages pile into a bell — no matter what the population looks like. Skewed incomes, bimodal wait times, anything with a finite \(\sigma\): the \(\bar{x}\)'s still go normal.

\(\bar{x} \approx N\!\left(\mu,\ \sigma^2/n\right)\)

The license conditions

Independence — a random sample. Size — \(n \geq 30\) as the rule of thumb for non-normal populations; if the population is already normal, any \(n\) gives an exactly normal pile of averages. Finite \(\sigma\).

\(n \geq 30\)

What it does NOT say

Your data never turns normal — 50 skewed incomes stay skewed at any \(n\); only the pile of averages goes bell. And no sample size rescues broken independence: a time series or repeated measures on the same people fails at \(n = 20\) and at \(n = 20{,}000\).

This is why the bell from last class is everywhere: even when individual values refuse it, their averages wear it. Watch the bell assemble out of a skewed population:

See it · C3

A skewed population, a bell of averages

Predict first: the population is right-skewed. At \(n = 2\), will the pile of averages be skewed or bell-shaped? Then I raise \(n\) to 30 and draw 100 samples — call the shape before it lands.

Figure 1: Sampling Distribution Builder — draw samples from the population (left) and watch the distribution of sample means grow on the right. Notice how the histogram becomes bell-shaped as more samples are added, regardless of the population shape.

Quick check · what goes normal?

Practice · is the bell licensed? · fresh numbers on demand

Core concept 4 · C4

What sample size buys — and what it stops buying

The \(\sqrt{n}\) in the denominator is the entire economics of survey design. It explains why Maya sampled 300 — and why she didn't sample 3,000.

The \(\sqrt{n}\) law

With \(\sigma = 1.3\) h: one student wanders \(\pm 1.3\); an average of 100 wanders \(\pm 0.13\); Maya's 300 wander \(\pm 0.075\). Every extra respondent tightens the average's aim.

\(\sigma_{\bar{x}} = \dfrac{\sigma}{\sqrt{n}}\)

Diminishing returns

To halve the SE you must quadruple \(n\) — doubling only buys a factor of \(\sqrt{2} \approx 1.41\). Going from 300 to 1,200 interviews would shave the wander from 0.075 h to 0.038 h: four times the work for one extra decimal of calm.

\(\sqrt{4n} = 2\sqrt{n}\)

Precision is detection

A small SE is what lets a real deviation stand out from sampling noise. The narrower the pile of averages, the harder it is for "we just got an unusual sample" to explain away a surprising \(\bar{x}\) — the idea the last stretch of today runs on.

The skeptic asked "why trust 300 of 48,000?" — notice the 48,000 never entered a formula. What matters is \(n\), not the fraction sampled. See the curve of diminishing returns:

See it · C4

The SE curve: steep gains first, then a long flat

Predict first: where does the curve flatten — before or after \(n = 100\)? Find the point where doubling the sample stops feeling worth it.

SE = 12.00
A hyperbolic decay curve showing SE = 12/√n. The orange marker traces the curve as you drag the slider. The curve flattens dramatically past n = 50, illustrating diminishing returns.
Figure: SE = σ/√n as a function of sample size. Drag the slider to see how SE falls quickly at first, then levels off — doubling n only halves SE if you quadruple it.

Practice · what sample size buys · fresh numbers on demand

Core concept 5 · C5

The pipeline, upgraded: one new denominator

Everything assembles here. The pile of averages is normal (C3), centered on \(\mu\) (C2), with spread \(\text{SE} = \sigma/\sqrt{n}\) (C2, C4) — so last class's entire z-pipeline works on averages verbatim, with exactly one edit.

Step 1 · The SE first

Before any z, compute the yardstick that actually fits an average: \(\text{SE} = \sigma/\sqrt{n}\). This step is the whole upgrade — everything after it is old news.

\(\text{SE} = \dfrac{\sigma}{\sqrt{n}}\)

Step 2 · The z, with the new denominator

Same bridge as last class, new footing: \(z = (\bar{x} - \mu)/\text{SE}\). Divide by \(\sigma\) instead and the z comes out \(\sqrt{n}\) times too small — the trap with a starring role today.

\(z = \dfrac{\bar{x} - \mu}{\text{SE}}\)

Step 3 · Then it's last class, verbatim

Read the left-tail entry, complement for "greater than", subtract for "between". If you can read the table for one value, you can read it for an average — the table never knows the difference.

\(P(Z < z)\)

Name the object — one value or an average of \(n\)? — and the object picks the denominator. See the pipeline as one picture, then watch me run a full verdict out loud:

See it · C5

The whole pipeline in one picture

Predict first: which single box in this pipeline is new since last class? Everything else you've already done.

Step 1
SE = σ / √n
Step 2
z = (x̄ − μ) / SE
Step 3 — Φ(z)
Mini standard normal N(0,1) curve. The orange shaded area to the left of the z marker represents Φ(z) = P(Z ≤ z).
Step 4 — Answers
P(x̄ < x̄₀)
P(x̄ > x̄₀)
Two-tail
Figure: A live four-step pipeline. Edit any input field to see all four steps update instantly. Default values are from Example 2. The shaded area in Step 3 represents Φ(z) — the left-tail probability.

I do · watch a full verdict

  1. 1 I do
  2. 2 We do
  3. 3 You do

Watch me decide every step out loud — name the object first (one value, or an average?), then SE, z, read, convert, verdict.

Consumer watch: "the 500-gram cereal boxes feel light"

Step 1 · Name the object

Readers tip Maya off; the plant's spec says fills are normal with \(\mu = 500\) g, \(\sigma = 8\) g. She weighs a random \(n = 64\) boxes: \(\bar{x} = 498\) g. That 498 is an average of 64, not one box — so its yardstick is the SE, not \(\sigma\). The question: \(P(\bar{x} \leq 498)\) if the machine is honest.

Step 2 · The yardstick for averages

\(\text{SE} = 8/\sqrt{64}\) \(= 1\) g. Averages of 64 boxes wander by about a gram — so 498 sits two whole yardsticks below spec. I expect a small area; writing that expectation down is the cheapest error alarm there is.

Step 3 · Cross the bridge

\(z = \dfrac{498 - 500}{1}\) \(= -2.00\). Had I divided by \(\sigma = 8\) instead: \(z = -0.25\) — an ordinary-looking blip. The wrong denominator doesn't just change a number; it changes the story.

Step 4 · Read, convert, verdict

Table: \(P(Z < 2.00) = 0.9772\), so \(P(Z < -2.00)\) \(= 1 - 0.9772 = 0.0228\). If the machine were honest, only about 2 samples in 100 would average this low. That's not proof — but it's printable evidence, and Maya has a story.

The habit: name the object (one value, or an average of \(n\)?), compute the SE, standardize, read and convert, then say the verdict in words.

We do · you finish it

  1. 1 I do
  2. 2 We do
  3. 3 You do

Same tip, smaller sample: you re-run the pipeline and supply the verdict — and watch the sample size flip it.

The plant pushes back: "you'd have seen that with any 16 boxes"

Suppose Maya had weighed only \(n = 16\) boxes and found the same \(\bar{x} = 498\) g. Re-run the pipeline: same spec (\(\mu = 500\), \(\sigma = 8\)), same question — \(P(\bar{x} \leq 498)\).

  1. Name the object: still an average — of 16 boxes now. New yardstick: \(\text{SE} = 8/\sqrt{16}\) \(= 2\) g.
  2. Cross the bridge: \(z = (498 - 500)/2\) \(= -1.00\). The table says \(P(Z < 1.00) = 0.8413\).
  3. Finish it: \(P(\bar{x} \leq 498) = ?\) — and does the verdict survive?

    \(P(\bar{x} \leq 498)\) \(= 1 - 0.8413 = 0.1587\) — about one honest sample in six averages this low. The same 498 g that was alarming at \(n = 64\) is unremarkable at \(n = 16\): sample size decides what you can detect. (That's the C4 card "precision is detection," now with teeth.)

Practice · the full pipeline, solo · fresh numbers on demand

  1. 1 I do
  2. 2 We do
  3. 3 You do

Now you run the whole pipeline solo — fresh numbers every re-roll, the table entries you need on screen, support at its minimum.

Which tool applies? · mixes earlier classes + today

Production task · answer the skeptic

Exit ticket · 60 seconds · anonymous

What Maya files

The skeptic answered — with a distribution

The objection was correct and quantifiable: \(\bar{x}\) is a random variable. Its distribution — the sampling distribution — centers exactly on the truth (\(\mu_{\bar{x}} = \mu\)), wanders only \(\text{SE} = \sigma/\sqrt{n}\), and is bell-shaped by the Central Limit Theorem once independence and size conditions hold.

For her survey: \(\text{SE} = 1.3/\sqrt{300} \approx 0.075\) h — a different 300 lands within about \(\pm 0.15\) h of the truth. And the cereal check showed the pipeline's teeth: the same 498 g is a 2-in-100 alarm at \(n = 64\) and a 1-in-6 shrug at \(n = 16\). Sample size decides what you can detect.

The one upgraded move: name the object first. One value → \(\sigma\). An average of \(n\) → \(\sigma/\sqrt{n}\). A count of trials → the binomial. After that, it's the same table, the same left-tail sentence, the same convert-before-you-report.

Maya can now say how far a different 300 could wander — but "within about ±0.15 h" is still hand-drawn. Next class makes it exact and printable: an interval around 6.8 with a stated warranty — the confidence interval, and with it the claim she's been chasing since the first survey: what all 48,000 students' average really is.

Vote counts on today's slides are simulated for rehearsal; live sessions show the room's real votes.