← Back to the Introduction to Statistics outline The Course Maker
← Introduction to Statistics outline
Week 5 · Lecture outline

Week 5 — Lecture Outline · Probability Foundations

Introduction to Statistics Generic evergreen edition

Course: Introduction to Statistics (18-week generic edition)
Objectives covered: Objective 4 — Apply probability rules, conditional probability, and random variables to quantify chance. (This week builds the probability-rules and conditional-probability portion; random variables complete the objective in Weeks 6–7.)
SLOs touched: A (reason quantitatively from data) · B (communicate results to a non-technical audience)
Meeting pattern: planned as 2 sessions × ~75 min = ~150 min. Segment minutes below total ~150; scale them to your own pattern.

Week at a Glance
The week's big question "What does 'a 1-in-6 chance' actually promise — and what does it never promise?"
By the end of the week, students can… (1) interpret a probability as a long-run relative frequency (and say what it does not promise about the next trial); (2) list a sample space and compute probabilities for equally likely outcomes; (3) apply the complement, addition (disjoint and general), and multiplication (independent events) rules — and say when each applies; (4) tell disjoint from independent; (5) read a conditional probability from a two-way table without swapping P(A | B) and P(B | A).
Key vocabulary random phenomenon, probability, law of large numbers, simulation, sample space, outcome, event, complement, disjoint (mutually exclusive) events, addition rule, general addition rule, independence, multiplication rule for independent events, conditional probability, two-way table, gambler's fallacy
Materials slides (Deck 5), the Week 5 chapter, the week's readings + video links, a spreadsheet (Google Sheets or Excel) for the =RANDBETWEEN simulation, the student's chatbot for the AI-critique moment and the tutorial
Timing note 8 segments, ~150 min total. Session 1 = Segments 1–4 (~75). Session 2 = Segments 5–8 (~75).

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

Hook. "Someone at your table rolls three sixes in a row. Be honest — what does your gut say about the fourth roll? 'The streak continues'? 'A six can't come again'? Both guts are wrong, and by the end of this week you'll know exactly why."

  • For four weeks we've described data we already had. But every claim we care about — "this poll is within 3 points," "this test is 95% reliable" — leans on a different question: what are the chances? This week we learn the machinery of chance itself.
  • "Probability is the grammar of uncertainty. Learn five rules and you can read it fluently."

The promise (write it on the board): "By the end of this week you can take any chance claim — a 1-in-6 die, a '5% defect rate,' a 'given that it failed…' — and compute with it correctly using five rules, without falling for the two oldest traps in gambling."

Why it matters line (memory hook): "Probability is a long-run promise, not a short-run guarantee."

Segment 2 — What Probability IS: Randomness & the Long Run (18 min)

Plain language first.
- A random phenomenon is one where individual outcomes are unpredictable but the long-run pattern is stable. You can't call one roll of a die; you can absolutely call what 10,000 rolls will look like.
- The probability of an event is the proportion of times it would occur in a very long run of repetitions. P(six) = 1/6 means: roll forever, and the fraction of sixes settles at one-sixth. It does not mean "one six in every six rolls."
- The law of large numbers: as the number of repetitions grows, the running proportion of an event settles toward its true probability. Short runs are wild; long runs are tame.

One fully worked example (do every step out loud).

Watching 1/6 emerge. A student simulates die rolls and tracks the running proportion of sixes:
- After 10 rolls: 3 sixes → 3 ÷ 10 = 0.300. (Nearly double the truth! Nothing is wrong.)
- After 100 rolls: 21 sixes → 21 ÷ 100 = 0.210.
- After 1,000 rolls: 172 sixes → 172 ÷ 1,000 = 0.172.
- The theoretical value: 1 ÷ 6 ≈ 0.1667. Each checkpoint is closer than the last (0.300 → 0.210 → 0.172 → …). That settling-down is the law of large numbers — and it's exactly what students will watch happen in this week's Data Lab with =RANDBETWEEN(1,6).

Land the key idea: probability describes the process, not the next result. A weather app's "70% rain" isn't a promise about your picnic; it's a statement about days like this one, in the long run.

Segment 3 — Sample Spaces, Events & Equally Likely Outcomes (22 min)

Plain language first.
- The sample space S is the list of every possible outcome of a random phenomenon. One die: S = {1, 2, 3, 4, 5, 6}. One card from a standard 52-card deck: 52 outcomes.
- An event is any collection of outcomes you care about ("roll an even number" = {2, 4, 6}).
- Probabilities live on a fixed scale: 0 ≤ P(A) ≤ 1, and the whole sample space has probability 1. A "probability" of 1.4 or −0.2 isn't a bold claim — it's an arithmetic error.
- When all outcomes are equally likely: P(A) = (number of outcomes in A) ÷ (total number of outcomes). Equally likely must be earned — a die is symmetric; a thumbtack landing point-up vs. point-down is not.

One fully worked example (build the grid).

Two dice, one question. Roll two fair dice (think: one red, one blue). Every (red, blue) pair is an outcome: 6 × 6 = 36 equally likely outcomes.
- Event: "the sum is 7." Count the ways: (1,6), (2,5), (3,4), (4,3), (5,2), (6,1) — 6 ways.
- P(sum is 7) = 6 ÷ 36 = 1/6 ≈ 0.167.
- Quick contrast with one card: P(heart) = 13 ÷ 52 = 1/4 = 0.25 — same recipe: count the favorable, divide by the total.

Notation lands after the idea: we write P(A) for "the probability of event A" — capital P, the event in parentheses. That's the entire notation load of the week until the conditional bar appears in Segment 6.

Segment 4 — The Complement & Addition Rules + Misconceptions + Quick Interaction (27 min) · Session 1 closes (~75)

The complement rule (plain language first).
- The complement of A ("not A") is everything in the sample space that isn't A. Since something must happen: P(not A) = 1 − P(A). The chance it happens and the chance it doesn't split 1 between them.

Worked example: a prize spinner has 10 equal sectors, 3 of them red. P(red) = 3/10 = 0.3, so P(not red) = 1 − 0.3 = 0.7. Ten seconds, no counting of seven sectors needed — that's the point of the rule.

The addition rules (plain language first).
- Two events are disjoint (mutually exclusive) if they can't both happen on the same trial. One card can't be both a seven and a king.
- Addition rule (disjoint): P(A or B) = P(A) + P(B). ("Or" in statistics always means A, B, or both.)
- General addition rule: P(A or B) = P(A) + P(B) − P(A and B). Subtract the overlap, because plain adding counts it twice.

One fully worked example (the double-count, exposed).

One card: P(heart or face card)? Hearts: 13. Face cards (J, Q, K): 12. Add: 13/52 + 12/52 = 25/52 — wrong, because the jack, queen, and king of hearts got counted twice.
- Overlap: heart AND face = 3 cards → 3/52.
- P(heart or face) = 13/52 + 12/52 − 3/52 = 22/52 = 11/26 ≈ 0.423.
- Memory hook: "OR adds — then subtracts the overlap."

Name the misconceptions out loud, then cure each:

  • ❌ "There are two outcomes, so it's 50/50."
    ✅ Cure: equally likely must be checked, not assumed. A thumbtack lands point-up or point-down — not 50/50. A skydive succeeds or fails — mercifully not 50/50. Count outcomes only when symmetry earns it.
  • ❌ "P(A or B) is always P(A) + P(B)."
    ✅ Cure: adding double-counts the overlap. Ask "can they both happen?" first; if yes, subtract P(A and B). (The heart/face example is the forever-cure.)
  • ❌ "A probability of 1.2 just means 'really likely.'"
    ✅ Cure: the scale is 0 to 1, always. An answer outside it isn't enthusiasm — it's a computation error, usually a double-count.
  • ❌ "P(six) = 1/6 means one six in every six rolls."
    ✅ Cure: long-run promise, not short-run guarantee. Six rolls can easily show zero sixes (or three).

Interaction — Think-Pair-Share (rapid-fire, ~10 min):
Put 6 quick items on a slide; solo 30 sec, compare 1 min, class votes. Items: (1) one card: P(spade)? (2) one roll: P(not a 5)? (3) one spin of a one-sector-at-a-time spinner: can "lands red" and "lands blue" both happen? (4) one roll: are "even" and "6" disjoint? (5) one card: P(king or red card)? (6) true or false — a probability of 1.15 is possible for a very likely event.
(Answers: 13/52 = 1/4 · 5/6 · no — disjoint · no — 6 is even, they overlap · 4/52 + 26/52 − 2/52 = 28/52 = 7/13 ≈ 0.538 · false — the scale stops at 1.)
Debrief item 5 — the double-count trap in the wild — and item 4, which quietly previews independence: "disjoint" is about overlap, not about influence.

Segment 5 — Independence & the Multiplication Rule (and the Gambler's Fallacy) (20 min) · Session 2 opens

Hook back in: "Last session: OR. This session: AND — and the single most expensive mistake in the history of gambling."

Plain language first.
- Two events are independent if knowing one happened doesn't change the probability of the other. Two spins of a spinner: what the first spin did tells you nothing about the second.
- Multiplication rule (independent events): P(A and B) = P(A) × P(B). "AND multiplies — when the trials don't talk to each other."

Worked example: that same 10-sector spinner (3 red). P(red on one spin) = 0.3. Two spins, independent: P(red then red) = 0.3 × 0.3 = 0.09. About a 9% chance — and notice it's smaller than either single event. "And" is a harder demand than "or."

Disjoint vs. independent (the classic mix-up — say it twice):
- Disjoint = can't both happen (an overlap fact). Independent = one tells you nothing about the other (an information fact).
- They are not synonyms — in fact, disjoint events are never independent (unless one is impossible): if "seven" and "king" are disjoint and I tell you the card is a seven, the probability it's a king just crashed to 0. Knowing A changed P(B) completely — that's maximal dependence.

The gambler's fallacy (name the trap):
- ❌ "No six in 20 rolls — a six is due." ✅ Cure: the die has no memory. Every roll, P(six) = 1/6, regardless of history. Independence is exactly the statement that history doesn't lean on the next trial.
- Reconcile it with the law of large numbers, because students feel a contradiction: the long run fixes proportions by swamping, not by compensating. A surplus of non-sixes isn't "paid back" by extra sixes later; it's diluted by thousands of ordinary rolls.

The famous failure (the story that sticks):

The Monte Carlo night, 1913. At a roulette table, black came up an astonishing 26 times in a row. As the streak grew, bettors piled ever-larger sums on red — reasoning that red was overwhelmingly "due" — and reportedly lost millions of francs to a wheel that had no idea what it had just done. Each spin: same probabilities, no memory.
The lesson, in one line: "Chance has no memory — and betting that it does has a price."

Segment 6 — Conditional Probability & Two-Way Tables (25 min)

Callback first: "Week 4: you read conditional distributions out of a two-way table — 'among the Adélie penguins, what fraction…' This week the same move gets its official name and symbol."

Plain language first.
- A conditional probability P(A | B) is the probability of A given that B happened. The vertical bar reads "given." Whatever follows the bar is your new world: shrink the table to B's row (or column), and re-count inside it.
- Formula (after the idea): P(A | B) = P(A and B) ÷ P(B). Rearranged, it's the general multiplication rule — P(A and B) = P(B) × P(A | B) — which collapses to Segment 5's rule exactly when A and B are independent, because then P(A | B) = P(A).

One fully worked example (the week's signature table — walk every cell).

The factory's end-of-day quality check. A small plant checks all 200 widgets made in one day on its two machines:

Defective Not defective Total
Machine A 6 114 120
Machine B 10 70 80
Total 16 184 200
  • P(defective) = 16 ÷ 200 = 0.08 — the plant-wide rate.
  • P(defective | Machine A) = 6 ÷ 120 = 0.05 — new world: A's row only.
  • P(defective | Machine B) = 10 ÷ 80 = 0.125 — B's widgets fail at 2.5× A's rate.
  • P(Machine B | defective) = 10 ÷ 16 = 0.625 — different question entirely: a defective widget turned up; how likely is it B's? New world: the defective column.
  • Formula check (same answer, no shrinking): P(defective | B) = P(defective and B) ÷ P(B) = (10/200) ÷ (80/200) = 0.05 ÷ 0.40 = 0.125. ✓
  • Independence check: is "defective" independent of "machine"? Compare P(defective | B) = 0.125 with P(defective) = 0.08. Not equal → not independent — the machine matters, and it's B that needs the maintenance visit.

Misconception + cure (the error headlines love):
- ❌ "P(defective | B) and P(B | defective) are the same thing."
✅ Cure: 0.125 ≠ 0.625. The first asks about B's output; the second asks about the defect pile. Swap the bar, change the question — always say the "given" world out loud in words before computing.

Segment 7 — Which Rule When? + the Streak Debate (18 min)

The decision guide (put it on one slide; students photograph the deck's toolkit slide):
- The question says "not" → complement: 1 − P(A).
- The question says "or" → check for overlap: disjoint? add. Overlap? add, then subtract P(A and B).
- The question says "and" across repetitions that don't influence each other → multiply.
- The question says "given," "among," "of the ones that…" → conditional: shrink your world, re-count.
- Sanity check on every answer: is it between 0 and 1? Does it make sense said aloud in words?

Rapid drill — name the rule, then compute (answers in parentheses):
1. A spinner lands on silver with probability 0.15. P(not silver)? (complement; 1 − 0.15 = 0.85)
2. One card: P(heart or spade)? (disjoint addition; 13/52 + 13/52 = 26/52 = 1/2)
3. Two rolls of a die: P(even then even)? (independent multiplication; 1/2 × 1/2 = 1/4)
4. From the factory table: P(defective | Machine A)? (conditional; 6/120 = 0.05)

Quick mini-debate (genuinely arguable, ~6 min): "It has rained three days in a row. One friend says a sunny day is 'due.' A farmer says another rainy day is MORE likely. Who's reasoning better?" Have students argue both sides, then surface the resolution: the "due" friend is running the gambler's fallacy — but the farmer isn't, because weather has memory (storm systems persist; tomorrow is not independent of today). Independence is a modeling choice that fits dice, spinners, and shuffled decks — machines without memory — and often fails for weather, skill, and streaky humans. The skill this week builds: asking "is independence actually plausible here?" before multiplying. (This is exactly the week's Discussion question — tease it.)

Segment 8 — Technology Workflow + AI-Critique, Callback & Hand-off (12 min) · Session 2 closes (~75)

Technology workflow — 500 die rolls in ten seconds (exact steps):
1. In A1:D1 type headers: roll, result, running sixes, running proportion.
2. A2 = 1; A3 = =A2+1; fill column A down to A501 (rolls 1–500).
3. B2 = =RANDBETWEEN(1,6); fill down to B501 — 500 virtual die rolls.
4. C2 = =COUNTIF($B$2:B2,6); fill down — sixes so far.
5. D2 = =C2/A2; fill down — the running proportion. Chart column D as a line chart and watch it thrash early, then settle toward 0.1667.
- Google Sheets and Excel: identical formulas. Any edit (or F9) re-rolls all 500 dice — that's the randomness working; copy → paste-values to freeze a run.

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

Paste this to your chatbot: "I rolled a virtual die 500 times and got 78 sixes. Is my die unfair? Also, my last 30 rolls included only 2 sixes — is a six more likely now?"
Then check its answers against this week's ideas. 78/500 = 0.156 is comfortably within ordinary chance variation for 500 rolls, and no, a six is not "more likely now" — chatbots sometimes hedge toward the gambler's fallacy to be agreeable, or declare a fair die "biased" from a gap that's pure noise. The tool drafts, you judge — and this week's Data Lab makes you do exactly this with your own 500 rolls.

Callback + tease:
- Callback: "Week 4 ended with a warning — a correlation isn't a cause. This week added the deeper machinery: chance itself follows rules, and the two-way tables you built in Week 4 are secretly probability machines."
- Tease next week: "Week 6 attaches payoffs to outcomes: what's a raffle ticket really worth? Random variables and expected value — probability starts earning money."

Hand-off (the week's work):
- Chapter 5 (the primary reading) — then Lecture Tutorial 5 (AI tutor, share-link submission) — the five rules, hands-on.
- Data Lab 5 (simulate 500 rolls; watch the law of large numbers happen) · Quiz 5 (end of week) · Discussion 5 (the lucky-streak question) · Assignment 5 (AI-coached).

Instructor FAQ — Common Stumbles

Student says / does Quick cure
"Two outcomes, so it's 50/50." Equally likely must be checked, not assumed. A thumbtack isn't a coin. Count outcomes only when symmetry earns it.
Adds P(A) + P(B) for overlapping events (gets 25/52 on the heart/face problem). "Can they both happen?" If yes, the overlap got counted twice — subtract P(A and B). Answers above 1 are the smoke alarm for this error.
"No six in 20 rolls — a six is due." The die has no memory: P(six) = 1/6 every roll. The long run fixes proportions by swamping, not compensating (Monte Carlo, 1913: 26 blacks in a row).
Uses disjoint and independent interchangeably. Disjoint = can't both happen (overlap fact). Independent = one tells you nothing about the other (information fact). Disjoint events are never independent — knowing one happened kills the other.
Swaps P(A|B) and P(B|A). Say the "given" world in words first. Factory table: P(defective | B) = 0.125 lives in B's row; P(B | defective) = 0.625 lives in the defective column. Different worlds, different denominators.
"My simulation gave 0.19, not 0.1667 — I did something wrong." Nothing's wrong — that's chance variation. For 500 rolls, roughly 0.13–0.20 is ordinary. More rolls tighten the band; that's the law of large numbers.
Reports a probability of 1.3 (or −0.2) without blinking. Install the reflex: every probability answer gets the 0-to-1 scale check before it leaves the page. Outside the scale = arithmetic error, usually a double-count.
"Why can't I always multiply for 'and'?" Multiplying plain P(A) × P(B) requires independence. If the events influence each other, you need the conditional version — P(A) × P(B | A) — or a two-way table, which handles it for you.

Scope flag

This outline stays within the probability-rules portion of Objective 4, at the depth inference needs (per the course profile: no combinatorics-heavy counting). The general multiplication rule P(A and B) = P(A) × P(B | A) appears once, as the conditional formula rearranged, but only its independent-events form is drilled. The 1913 Monte Carlo story and the weather-streak mini-debate are added context (not strictly required by the objective) — kept because they make the gambler's-fallacy and independence-as-a-choice lessons stick; cut them for a leaner session.