Poisson approximation to the binomial

Poisson approximation to the binomial

Spec 1.3 · Edexcel International A Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • For X∼B(n,p)X\sim B(n,p) with nn large and pp small, X≈Po(np)X\approx\text{Po}(np).
  • A guide: n>50n>50 and p<0.1p<0.1.
  • The parameter is λ=np\lambda=np, the binomial mean.
  • The fit is poor when pp is not small, because the variance np(1−p)np(1-p) is then well below the mean.

Why approximate

When pp is small the binomial variance is close to its mean, just like a Poisson.

Finding P(X=r)=(nr)pr(1−p)n−rP(X=r)=\binom nr p^r(1-p)^{n-r} for large nn is awkward, and binomial tables do not cover large nn. When nn is large and pp is small, the binomial distribution behaves very like a Poisson distribution, which is easier to use.

When pp is small, 1−p≈11-p\approx1, so np(1−p)np(1-p) is close to the mean npnp. A Poisson has variance equal to its mean.

00.050.10.150.20.250.0120.0120123456rProbability
  • B(1000, 0.002)
  • Po(2)
B(1000, 0.002) and Po(2) give almost identical probabilities, because p is small and the variance (1.996) is close to the mean (2).

Binomial B(n, p)

Mean:
np
Variance:
np(1 − p)
Variance close to mean when:
p is small

Poisson Po(λ)

Mean:
λ
Variance:
λ
Variance close to mean when:
always

X∼B(100,0.01)X\sim B(100,0.01). What is the variance?

The approximation

If nn is large and pp is small, then X≈Po(np)X\approx\text{Po}(np).

As a guide, use it when n>50n>50 and p<0.1p<0.1; the smaller pp and larger nn, the better the fit. The parameter is λ=np\lambda=np, the binomial mean.

00.050.10.150.20.00690.0081012345678rProbability
  • B(60, 0.05)
  • Po(3)
B(60, 0.05) with n > 50 and p < 0.1 is close to Po(3), where λ = np = 3.
  • B(n, p), n large, p small≈Po(np)\approx\text{Po}(np)

Worked example

State a Poisson approximation for X∼B(80,0.05)X\sim B(80,0.05).

X∼B(300,0.01)X\sim B(300,0.01). Which Poisson approximation is correct?

Using the approximation

Once you have Po(np)\text{Po}(np), work exactly as for any Poisson variable.

Use Poisson tables or the formula, rewriting inequalities in terms of P(X≤r)P(X\le r). State the new distribution first.

For combined questions, such as 10 independent boxes, find the single-box probability with the approximation and then use the binomial or product rule for the boxes.

00.050.10.150.2012345678rProbability
  • B(200, 0.02)
  • Po(4)
B(200,0.02)B(200,0.02) and Po(4)\text{Po}(4) give almost the same probabilities.

Worked example

X∼B(200,0.02)X\sim B(200,0.02). Use a Poisson approximation to find P(X≤2)P(X\le2) and P(X>5)P(X>5).

After approximating B(200,0.02)B(200,0.02) by Po(4)\text{Po}(4), which expression gives P(X<3)P(X<3)?

Judging accuracy

It is good when nn is large and pp is small, and poor otherwise.

The approximation is a model, so the answer is not exact. For B(1000,0.002)B(1000,0.002) the approximate P(X≥4)P(X\ge4) is 0.14290.1429 and the exact value is 0.14270.1427.

It gets worse when pp is larger or nn is small. A binomial variable can never exceed nn, while a Poisson variable has no upper limit; this is harmless when nn is large and pp is small.

00.050.10.150.20.250.0060.0106012345678rProbability
  • B(10, 0.4)
  • Po(4)
B(10, 0.4) and Po(4) do not match: p is not small and n is small, so the approximation is poor.

B(1000, 0.002)

Mean:
2
Variance:
1.996
Good approximation?:
Yes: variance close to mean

B(10, 0.4)

Mean:
4
Variance:
2.4
Good approximation?:
No: variance well below mean

Which binomial is best approximated by a Poisson distribution?

Try an exam question

X∼B(150,0.02)X\sim B(150,0.02). (a) Explain why a Poisson distribution is a suitable approximation. (b) Use it to find P(X≤2)P(X\le2).

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Which Poisson distribution is the best approximation to the distribution of XX?

1 mark, marked against the mark scheme

Revision notes