Lesson 6 of 8 · 18 min

Bernoulli trials and binomial probabilities

Recognise when the number of successes in repeated trials is binomial, derive the binomial probability formula from a tree diagram, and use it by hand.

What you will be able to do

  • State the four conditions for a binomial random variable and check them in context.
  • Explain where P(X=r)=(nr)pr(1−p)n−rP(X = r) = \binom{n}{r} p^r (1 - p)^{n - r} comes from.
  • Calculate binomial probabilities by hand, using X∼Bin(n,p)X \sim \mathrm{Bin}(n, p) notation.

You haven't studied for a quiz of three multiple-choice questions, each with four options, so you guess every answer. How many will you get right? Each guess is a Bernoulli trial with p=14p = \frac{1}{4}, and you are counting the successes over three trials. The tree shows every possibility.

Interactive

Guessing three questions

Tick every path with exactly two right answers. What do their products have in common?

Three paths have exactly two right answers: RRW, RWR and WRR. Each has two factors of 14\frac{1}{4} and one of 34\frac{3}{4}, so each has the same probability, (14)2(34)\left(\frac{1}{4}\right)^2 \left(\frac{3}{4}\right). The order doesn't change the product, so

P(X=2)=3×(14)2(34)=964.P(X = 2) = 3 \times \left(\tfrac{1}{4}\right)^2 \left(\tfrac{3}{4}\right) = \tfrac{9}{64}.

The 3 is the number of ways to choose which two of the three questions are right: (32)=3\binom{3}{2} = 3.

Using the tree, find the probability of getting none of the three questions right.

Reads asYour answer shows here as maths.

The four conditions for a binomial random variable

  1. A fixed number of trials, nn

    The number of trials is decided in advance (three questions, 20 seeds, 10 kicks).

  2. Two outcomes on each trial

    Each trial is a success or a failure: a Bernoulli trial.

  3. The same probability of success, pp, on every trial

    It doesn't drift from trial to trial.

  4. Independent trials

    The outcome of one trial doesn't change the probabilities for any other.

Key idea

The binomial distribution

If XX is the number of successes rr in nn independent Bernoulli trials, each with probability of success pp, then XX has a binomial distribution, written X∼Bin(n,p)X \sim \mathrm{Bin}(n, p), and

P(X=r)=(nr)pr(1−p)n−r,r=0,1,2,…,n.P(X = r) = \binom{n}{r} p^r (1 - p)^{n - r}, \qquad r = 0, 1, 2, \ldots, n.

pr(1−p)n−rp^r (1 - p)^{n - r} is the probability of one particular order of rr successes and n−rn - r failures, and (nr)\binom{n}{r} counts the orders.

Why the probabilities add to 1

By the binomial theorem from Unit 1,

∑r=0n(nr)pr(1−p)n−r=(p+(1−p))n=1n=1.\sum_{r=0}^{n} \binom{n}{r} p^r (1 - p)^{n - r} = \big(p + (1 - p)\big)^n = 1^n = 1.

That is why it is called the binomial distribution: its probabilities are the terms of the expansion of (p+q)n(p + q)^n with q=1−pq = 1 - p.

Worked example

Binomial probabilities by hand

  • Simple familiar
  • Technology free

In a game, a player wins a round with probability 13\frac{1}{3}, independently of other rounds. A player plays 5 rounds. Let XX be the number of rounds won.

(a) State the distribution of XX.

(b) Determine P(X=2)P(X = 2).

(c) Determine the probability that the player wins at most one round.

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Check the binomial conditions and identify nn and pp.

  2. Use P(X=r)=(nr)pr(1−p)n−rP(X = r) = \binom{n}{r}p^r(1 - p)^{n - r} for each value needed.

  3. Add the values included in the event.

Watch out for

  • Leaving out (52)\binom{5}{2}.

  • Swapping the powers: (13)3(23)2\left(\frac{1}{3}\right)^3\left(\frac{2}{3}\right)^2 is for 3 wins.

X∼Bin(4,0.2)X \sim \mathrm{Bin}(4, 0.2). Find P(X=0)P(X = 0).

A class has 12 girls and 10 boys. Four students are chosen at random, one after another, to form a committee. Why is the number of girls chosen not binomial?

Choose one answer

Note

Large populations

Choosing 4 people from a population of a million without replacement barely changes pp, so a binomial model is a very good approximation. The trouble comes when the sample is a sizeable fraction of the group it is drawn from.

Common mistake

Forgetting the binomial coefficient

pr(1−p)n−rp^r(1 - p)^{n - r} on its own is the probability of one order, such as SSF. Multiply by (nr)\binom{n}{r} to count all the orders. A quick check: if your P(X=1)P(X = 1) for Bin(5,0.5)\mathrm{Bin}(5, 0.5) is 132\frac{1}{32}, you've forgotten the (51)=5\binom{5}{1} = 5.

Exam tip

State the model

Before any calculation, write "Let XX be the number of … . Then X∼Bin(n,p)X \sim \mathrm{Bin}(n, p) with n=…n = \ldots and p=…p = \ldots." In a modelling question, also say why the conditions hold, or which one is doubtful.

Summary

  • A binomial random variable counts the successes in nn independent Bernoulli trials with constant pp.

  • Check the four conditions: fixed nn, two outcomes, constant pp, independence.

  • X∼Bin(n,p)X \sim \mathrm{Bin}(n, p) has P(X=r)=(nr)pr(1−p)n−rP(X = r) = \binom{n}{r} p^r (1 - p)^{n - r} for r=0,1,…,nr = 0, 1, \ldots, n.

  • (nr)\binom{n}{r} counts the orders; pr(1−p)n−rp^r(1 - p)^{n - r} is the probability of each order.

  • Drawing without replacement from a small group is not binomial.

Finished the lesson?

Try the 3 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 6 of 80% read

Save your progress with a free account