Lessons and exercises follow the school system and stay in its language.Les leçons et les exercices suivent le système scolaire et restent dans sa langue.Lektionen und Übungen folgen dem Schulsystem und bleiben in seiner Sprache.

Probability and decisions

Print the chapter

Every lesson of the chapter, ready for paper. Exercise sets are lettered A, B, C… and answered at the end.

To save a PDF, choose “Save as PDF” in the print dialog.

1. Discrete random variables and their distributions

Warm-up

Three quick questions from earlier lessons.

  1. Two fair dice are rolled. Find the probability that the sum is 2.
  2. A fair die is rolled 2 times. Find the probability of at least one six.
  3. The probability that a flight is late is 0.78. Find the probability that it is not late.

Discover

X is the number of heads in two coin tosses. Its values are 0, 1, 2. Which list gives their probabilities?

  1. 14,12,14\tfrac{1}{4}, \tfrac{1}{2}, \tfrac{1}{4}
  2. 13,13,13\tfrac{1}{3}, \tfrac{1}{3}, \tfrac{1}{3}

Key idea

A random variable assigns a number to each outcome.

A discrete random variable XX has a list of values with probabilities P(X=x)P(X = x). Every probability is between 0 and 1, and they sum to 1. Probabilities of events such as X≥2X \ge 2 add the matching values.

Worked example

P(X=0)=0.2P(X = 0) = 0.2, P(X=1)=0.5P(X = 1) = 0.5, XX takes only 0, 1, 2.

P(X=2)=1−0.2−0.5=0.3P(X = 2) = 1 - 0.2 - 0.5 = 0.3
Sum to 1.
P(X≥1)=0.8P(X \ge 1) = 0.8
Everything but 0.

Your turn, guided

Your turn: P(X=0)=0.1P(X = 0) = 0.1, P(X=1)=0.6P(X = 1) = 0.6. Find P(X=2)P(X = 2).

0.30.3
1 − 0.1 − 0.6.

Practice ★

  1. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 29/100 and P(X = 1) = 13/100. Find P(X = 2).
  2. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 14/100 and P(X = 1) = 18/100. Find P(X = 2).
  1. Which list could be the probabilities of a random variable with three values?
    1. 0.1, 0.5 and 4/10
    2. 0.5, 0.5 and 0.1
    3. 0.8, 0.4 and −0.2

Watch out!

Practice ★★

  1. X takes the values 0, 1, 2 with P(X = 0) = 25/100 and P(X = 1) = 15/100. Find P(X ≥ 1).
  2. X takes the values 0, 1, 2 with P(X = 0) = 34/100 and P(X = 1) = 15/100. Find P(X ≥ 1).

Exit ticket

Three questions to check you have got it.

  1. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 29/100 and P(X = 1) = 13/100. Find P(X = 2).
  2. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 14/100 and P(X = 1) = 18/100. Find P(X = 2).
  3. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 18/100 and P(X = 1) = 23/100. Find P(X = 2).

Summary card

Discrete distributions

  • Values with probabilities in [0,1][0, 1] summing to 1.
  • Add probabilities for events.

Watch out: values are not automatically equally likely.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. X is equally likely to be any whole number from 1 to 9. Find P(X ≥ 3).

2. Expected value and standard deviation of a random variable

Warm-up

Three quick questions from earlier lessons.

  1. A random variable X takes the values 0, 1 and 2, with P(X = 0) = 29/100 and P(X = 1) = 13/100. Find P(X = 2).
  2. X takes the values 0, 1, 2 with P(X = 0) = 25/100 and P(X = 1) = 15/100. Find P(X ≥ 1).
  3. Which list could be the probabilities of a random variable with three values?
    1. 0.1, 0.5 and 4/10
    2. 0.5, 0.5 and 0.1
    3. 0.8, 0.4 and −0.2

Discover

A raffle ticket wins 10 dollars with probability 0.1 and nothing otherwise. What is the average win per ticket over many tickets?

  1. 1 dollar
  2. 5 dollars

Key idea

The mean and SD of a random variable.

E(X)=μX=∑x P(x)E(X) = \mu_X = \sum x\,P(x), the long-run average. Var(X)=∑(x−μ)2P(x)\text{Var}(X) = \sum (x - \mu)^2 P(x) and SD(X)=Var(X)\text{SD}(X) = \sqrt{\text{Var}(X)}. For XX equal to 1 with probability pp and 0 otherwise: μ=p\mu = p, Var=p(1−p)\text{Var} = p(1 - p).

Worked example

XX: 0, 1, 2 with probabilities 0.3, 0.4, 0.3.

E(X)=0+0.4+0.6=1E(X) = 0 + 0.4 + 0.6 = 1
Σ x P(x).
Var(X)=0.3+0+0.3=0.6\text{Var}(X) = 0.3 + 0 + 0.3 = 0.6
Σ (x − 1)² P(x).
SD(X)=0.6≈0.77\text{SD}(X) = \sqrt{0.6} \approx 0.77
Square root.

Your turn, guided

Your turn: XX = 0, 5 with probabilities 0.6, 0.4. E(X)E(X).

22
5 × 0.4.

Practice ★

  1. X takes the values 0, 1, 2 with probabilities 51/100, 15/100 and 34/100. Find E(X).
  2. X takes the values 0, 1, 2 with probabilities 56/100, 25/100 and 19/100. Find E(X).
  1. A game costs 1 dollar. You win 4 dollars (net) with probability 1/3; otherwise you lose your dollar. Find your expected net gain, in dollars.

Watch out!

Practice ★★

  1. X = 1 with probability 7/10 and X = 0 otherwise. Find the standard deviation of X.
  2. X = 1 with probability 9/10 and X = 0 otherwise. Find the standard deviation of X.

Exit ticket

Three questions to check you have got it.

  1. X takes the values 0, 1, 2 with probabilities 51/100, 15/100 and 34/100. Find E(X).
  2. X takes the values 0, 1, 2 with probabilities 56/100, 25/100 and 19/100. Find E(X).
  3. X takes the values 0, 1, 2 with probabilities 77/100, 10/100 and 13/100. Find E(X).

Summary card

Expected value and SD

E(X)=∑xP(x)E(X) = \sum x P(x); SD=∑(x−μ)2P(x)\text{SD} = \sqrt{\sum (x - \mu)^2 P(x)}.

Watch out: weight by probabilities; include losses.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. A game costs 1 dollar. With probability 15\dfrac{1}{ 5 } you win ww dollars net; otherwise your dollar is lost. For which ww is the game fair (expected value 0)?

3. Long-run frequency and simulation

Warm-up

Three quick questions from earlier lessons.

  1. Two fair dice are rolled. Find the probability that the sum is 2.
  2. A fair die is rolled 2 times. Find the probability of at least one six.
  3. The probability that a flight is late is 0.78. Find the probability that it is not late.

Discover

A fair coin gave 7 heads in 10 tosses. If you toss it 10,000 times, what do you expect?

  1. A proportion of heads close to 0.5
  2. More tails than heads, to make up for the 7 heads

Run 10 tosses a few times, then 1,000. What happens to the proportion of heads?

  1. It gets close to 0.5
  2. It swings further from 0.5 to make up for streaks

Key idea

Probability is a long-run relative frequency.

Law of large numbers: as a random process is repeated many times, the proportion of times an event occurs gets close to its probability. Independent trials have no memory. A simulation imitates a random process with random numbers to estimate probabilities.

Worked example

Simulate a 30% chance with random digits.

Let 0, 1, 2 mean "success"
Each digit has probability 0.1.
Read 1,000 random digits; count successes
Repeat many times.
About 300 successes expected
Relative frequency.

Your turn, guided

Your turn: how many digits for a 60% chance?

66
0.6 ÷ 0.1.

Practice ★

  1. A fair coin lands heads 7 times in a row. What is the probability of heads on the next toss?
    1. 1/2: the coin has no memory
    2. Less than 1/2: tails is due
    3. More than 1/2: heads is on a streak
  1. What does it mean that the probability of rain tomorrow is 0.7?
    1. On many days like tomorrow, it rains on about 70% of them
    2. It will rain for 70% of the day
    3. It will rain if the probability is above 0.5

Watch out!

Practice ★★

  1. To simulate a free throw made with probability 0.6 using random digits 0–9, how many of the ten digits should stand for 'made'?
  2. To simulate a free throw made with probability 0.4 using random digits 0–9, how many of the ten digits should stand for 'made'?

Exit ticket

Three questions to check you have got it.

  1. A fair coin lands heads 7 times in a row. What is the probability of heads on the next toss?
    1. 1/2: the coin has no memory
    2. Less than 1/2: tails is due
    3. More than 1/2: heads is on a streak
  2. A fair coin lands heads 5 times in a row. What is the probability of heads on the next toss?
    1. 1/2: the coin has no memory
    2. Less than 1/2: tails is due
    3. More than 1/2: heads is on a streak
  3. A fair coin lands heads 10 times in a row. What is the probability of heads on the next toss?
    1. 1/2: the coin has no memory
    2. Less than 1/2: tails is due
    3. More than 1/2: heads is on a streak

Summary card

Long run

  • Probability = long-run relative frequency.
  • Simulate with random numbers.

Watch out: no memory, no "due" outcomes.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. A fair die is rolled 228 times. About how many sixes do you expect?

4. Tree diagrams and reverse conditioning

Warm-up

Three quick questions from earlier lessons.

  1. Two independent machines fail with probabilities 0.2 and 0.2. Find the probability that both fail.
  2. P(A) = 3/10, P(B) = 6/10 and P(A and B) = 18/100. Are A and B independent?
    1. Yes: P(A and B) = P(A)·P(B)
    2. No, because P(A and B) > 0
    3. It cannot be decided
  3. A bag has 3 red and 3 blue marbles. Two are drawn without replacement. Find P(both red).

Discover

1% of people have a condition. A test is positive for 95% of them, and for 5% of healthy people. Of 10,000 people, about how many test positive in total?

  1. About 590
  2. About 5,000

Key idea

A tree organizes conditional probabilities.

First branches: P(A)P(A) and P(not A)P(\text{not } A); second branches: conditional probabilities. Multiply along a path; add the paths that end in BB to get P(B)P(B). Then P(A∣B)=P(A and B)P(B)P(A \mid B) = \dfrac{P(A \text{ and } B)}{P(B)} (reverse conditioning, Bayes' rule).

Worked example

With the numbers above: P(condition | positive).

0.01×0.95=0.00950.01 \times 0.95 = 0.0095
Path: condition, positive.
P(+)=0.0095+0.99×0.05=0.059P(+) = 0.0095 + 0.99 \times 0.05 = 0.059
Both paths to positive.
P(condition∣+)=0.00950.059≈0.16P(\text{condition} \mid +) = \tfrac{0.0095}{0.059} \approx 0.16
Divide.

Your turn, guided

Your turn: P(A) = 0.2, P(B | A) = 0.5, P(B | not A) = 0.25. Find P(B).

0.30.3
0.2 × 0.5 + 0.8 × 0.25.

Practice ★

  1. 12% of a population has a condition. A test is positive for 86% of people with the condition and for 11% of people without it. Find P(positive).
  2. 14% of a population has a condition. A test is positive for 94% of people with the condition and for 8% of people without it. Find P(positive).
  1. 22% of a population has a condition. A test is positive for 81% of people with the condition and for 12% of people without it. Find P(condition | positive).

Watch out!

Practice ★★

  1. A disease affects 1 in 1,000 people. A test is 93% accurate for both sick and healthy people. Your test is positive. What is most likely?
    1. You are still more likely healthy: false positives from the many healthy people outnumber the true positives
    2. You are about 93% likely to be sick
    3. You are certainly sick

Exit ticket

Three questions to check you have got it.

  1. 12% of a population has a condition. A test is positive for 86% of people with the condition and for 11% of people without it. Find P(positive).
  2. 14% of a population has a condition. A test is positive for 94% of people with the condition and for 8% of people without it. Find P(positive).
  3. 28% of a population has a condition. A test is positive for 95% of people with the condition and for 14% of people without it. Find P(positive).

Summary card

Trees

  • Multiply along paths; add paths to the same end.
  • P(A∣B)=P(A and B)/P(B)P(A \mid B) = P(A \text{ and } B)/P(B).

Watch out: with rare events, most positives can be false.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. 16% of a population has a condition. A test is positive for 93% of people with it and for 13% of people without it. Your test is negative. Find the probability that you do not have the condition.

5. Waiting for a first success

Discover

You roll a fair die until you get a six. On average, how many rolls does it take?

  1. 6 rolls
  2. 3.5 rolls

Key idea

Waiting for a first success.

If each try succeeds with probability pp, independently, the expected number of tries until the first success is

E=1pE = \dfrac{1}{p}.

Worked example

A spinner lands on red with probability 14\tfrac{1}{4}. How many spins do you expect until the first red?

E=11/4E = \dfrac{1}{1/4}
p=14p = \tfrac{1}{4}.
E=4E = 4 spins
11/4=4\dfrac{1}{1/4} = 4.

Your turn, guided

Your turn: each try succeeds with probability 110\tfrac{1}{10}. How many tries do you expect?

1010
11/10\dfrac{1}{1/10}.

Practice ★

  1. Each try succeeds with probability 111\tfrac{1}{ 11 }, independently. How many tries do you expect until the first success?
  2. Each try succeeds with probability 115\tfrac{1}{ 15 }, independently. How many tries do you expect until the first success?
  3. Each try succeeds with probability 12\tfrac{1}{ 2 }, independently. How many tries do you expect until the first success?

Practice ★★

  1. A machine part fails on a given day with probability 160\tfrac{ 1 }{ 60 }. On average, how many days pass until its first failure?
  2. A machine part fails on a given day with probability 150\tfrac{ 1 }{ 50 }. On average, how many days pass until its first failure?

Exit ticket

Two questions to check.

  1. Each try succeeds with probability 111\tfrac{1}{ 11 }, independently. How many tries do you expect until the first success?
  2. Each try succeeds with probability 115\tfrac{1}{ 15 }, independently. How many tries do you expect until the first success?

Summary card

Waiting time

  • Expected waiting time =1p= \dfrac{1}{p}.
  • Rare events: long waits on average.

Engineers use it for the expected time between failures of a machine.

6. History: Bernoulli, de Moivre and the law of large numbers

Discover

Toss a fair coin 10 times: you may well get 7 heads. Toss it 10,000 times: the share of heads is almost always very close to 12\tfrac{1}{2}. Who first proved that this must happen?

  1. Jacob Bernoulli, with his law of large numbers
  2. Nobody: it is only an observation

Key idea

From one throw to many.

17001710172017301740175017601713Bernoulli1713Bernoulli
  1. Step 117001710172017301740175017601713Bernoulli1713BernoulliJacob Bernoulli 1655–1705Niklaus Bernoulli (1662-1716), 1687. Public domain1713: Jacob Bernoulli's Ars Conjectandi is printed, eight years after his death. It proves the law of large numbers.
  2. Step 217001710172017301740175017601713Bernoulli1718de Moivre1713Bernoulli1718de MoivreAbraham de Moivre 1667–1754Joseph Highmore, 1736. Public domain1718: de Moivre's book The Doctrine of Chances.
  3. Step 317001710172017301740175017601713Bernoulli1718de Moivre1733de Moivre1713Bernoulli1718de Moivre1733de MoivreAbraham de Moivre 1667–1754Joseph Highmore, 1736. Public domain13 November 1733: de Moivre approximates the binomial distribution by the bell curve.

Jacob Bernoulli's Ars Conjectandi (printed in 1713, eight years after his death) contains the law of large numbers. On 13 November 1733 Abraham de Moivre showed that for many trials the binomial distribution looks like the bell curve, today's normal distribution.

Worked example

With nn tosses of a fair coin, the number of heads XX has mean n2\tfrac{n}{2} and standard deviation n2\tfrac{\sqrt{n}}{2}.

mean 50, standard deviation 5
For n=100n = 100:
0.0050.005: very close to 12\tfrac{1}{2}
For n=10 000n = 10\,000: the share of heads has standard deviation 5010 000\tfrac{50}{10\,000}.

Practice ★

  1. What does Jacob Bernoulli's law of large numbers (printed in 1713) say?
    1. Over many trials, the frequency gets close to the probability
    2. After many tails, heads becomes more likely
    3. Large numbers are more likely than small ones
  1. A fair coin is tossed 144 times. With σ=n2\sigma = \tfrac{\sqrt{n}}{2}, what is the standard deviation of the number of heads?
  2. A fair coin is tossed 196 times. With σ=n2\sigma = \tfrac{\sqrt{n}}{2}, what is the standard deviation of the number of heads?
  1. Ars Conjectandi was printed in 1713, eight years after Jacob Bernoulli's death. In which year did he die? Add 4 to that year.

Summary card

Random variables today

  • Polls: a sample of 1,000 people gives a share close to the true one.
  • Insurance and casinos rely on the law of large numbers.
  • Quality control counts faulty items with the binomial distribution.

Sources: MacTutor biographies of Jacob Bernoulli and Abraham de Moivre (see "Sources and credits").