Mean, variance and Poisson approximation to the binomial

Mean, variance and Poisson approximation to the binomial

Spec 2.2–3 · Edexcel A-Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • Binomial B(n,p)B(n,p): E(X)=npE(X)=np and Var(X)=np(1−p)\text{Var}(X)=np(1-p), so the variance is smaller than the mean.
  • Poisson Po(λ)\text{Po}(\lambda): E(Y)=Var(Y)=λE(Y)=\text{Var}(Y)=\lambda.
  • From a binomial mean and variance: VarE=1−p\frac{\text{Var}}{E}=1-p, then n=Epn=\frac{E}{p}.
  • When nn is large and pp is small, B(n,p)≈Po(np)B(n,p)\approx\text{Po}(np).
  • The approximation overestimates the spread, so it is unsuitable when pp is not small.

Binomial mean and variance

For X∼B(n,p)X\sim B(n,p), the mean is npnp and the variance is np(1−p)np(1-p).

If X∼B(n,p)X\sim B(n,p), with nn independent trials each with success probability pp, then E(X)=npE(X)=np and Var(X)=np(1−p)\text{Var}(X)=np(1-p). Derivations are not required.

The variance is always smaller than the mean, because it is the mean multiplied by 1−p1-p. The standard deviation is the square root of the variance.

Do not use Var(X)=np\text{Var}(X)=np for a binomial: only the Poisson variance equals the mean.

0246810p = 0.1p = 0.25p = 0.5pValue
  • Mean
  • Variance
Mean np and variance np(1−p) for B(20, p). The variance is always smaller than the mean.
  • E(X)E(X)npnp
  • Var(X)\text{Var}(X)np(1−p)np(1-p)

Worked example

X∼B(40,0.15)X\sim B(40,0.15). Find the mean, variance and standard deviation.

X∼B(50,0.2)X\sim B(50,0.2). What is Var(X)\text{Var}(X)?

Poisson mean and variance

Both equal λ\lambda.

If Y∼Po(λ)Y\sim\text{Po}(\lambda) then E(Y)=λE(Y)=\lambda and Var(Y)=λ\text{Var}(Y)=\lambda. The mean and variance are equal. Derivations are not required.

If Y∼Po(7)Y\sim\text{Po}(7) the variance is 77 and the standard deviation is 7=2.65\sqrt7=2.65.

This equality is a useful check: if a data set has a mean and variance that are very different, a Poisson model is probably unsuitable. A binomial has variance less than its mean. Compare the sample mean and variance as a quick test, but also consider the context.

0246B(20, 0.35)Po(7)DistributionValue
  • Mean
  • Variance
B(20, 0.35) and Po(7) share a mean of 7, but only the Poisson variance equals its mean.

Binomial B(n, p)

Mean:
npnp
Variance:
np(1−p)np(1-p)
Variance compared with mean:
Smaller

Poisson Po(λ)

Mean:
λ\lambda
Variance:
λ\lambda
Variance compared with mean:
Equal

Y∼Po(7)Y\sim\text{Po}(7). What is the standard deviation of YY?

Finding n and p

Divide variance by mean to get 1−p1-p, then find n=E(X)pn=\frac{E(X)}{p}.

If you know E(X)E(X) and Var(X)\text{Var}(X) for a binomial variable, divide to find 1−p1-p: Var(X)E(X)=np(1−p)np=1−p\frac{\text{Var}(X)}{E(X)}=\frac{np(1-p)}{np}=1-p.

Then find nn from n=E(X)pn=\frac{E(X)}{p}.

Dividing mean by variance and using the result as pp is a common mistake: the ratio variancemean\frac{\text{variance}}{\text{mean}} is 1−p1-p. Always check that pp lies between 0 and 1 and that nn is a positive integer.

  1. 1

    Divide variance by mean

    9 ÷ 12 = 0.75 = 1 − p

  2. 2

    Find p

    p = 1 − 0.75 = 0.25

  3. 3

    Find n

    n = mean ÷ p = 12 ÷ 0.25 = 48

Worked pattern with a mean of 12 and a variance of 9.
  • variancemean\dfrac{\text{variance}}{\text{mean}}1−p1-p
  • nnmeanp\dfrac{\text{mean}}{p}

Worked example

A binomial variable has mean 1212 and variance 9.69.6. Find nn and pp.

A binomial variable has mean 2020 and variance 1515. Find nn and pp.

Poisson approximation

When nn is large and pp is small, B(n,p)≈Po(np)B(n,p)\approx\text{Po}(np).

When nn is large and pp is small, B(n,p)B(n,p) can be approximated by Po(np)\text{Po}(np) (the Poisson approximation). Derivations are not required.

It works because np(1−p)≈npnp(1-p)\approx np when pp is small, so the binomial mean and variance are nearly equal, as for a Poisson.

As a guide, nn should be at least about 50 and pp no more than about 0.10.1. State both conditions and give the new parameter λ=np\lambda=np.

00.050.10.150.2k = 0k = 1k = 2k = 3k = 4k = 5k = 6kProbability
  • B(100, 0.03)
  • Po(3)
P(X = k) for B(100, 0.03) and Po(3): the bars are almost identical when n is large and p is small.
  • nn large, pp smallB(n,p)≈Po(np)B(n,p)\approx\text{Po}(np)

Worked example

X∼B(500,0.004)X\sim B(500,0.004). Use a Poisson approximation to find P(X≥4)P(X\geq4).

X∼B(200,0.02)X\sim B(200,0.02) is approximated by a Poisson distribution. Which one?

Assessing the approximation

It keeps the mean but overestimates the spread, so it is poor when pp is not small.

The approximation keeps the mean npnp but replaces the variance np(1−p)np(1-p) by npnp, so it overestimates the spread, by a factor of 11−p\frac{1}{1-p}. For p=0.004p=0.004 the effect is tiny.

For B(20,0.3)B(20,0.3) the mean is 6 and the variance 4.2, while Po(6)\text{Po}(6) has variance 6. Also P(Y≤3)P(Y\leq3) is 0.1070.107 exactly but 0.1510.151 under the approximation, about 41% too large.

Conclude the approximation is unsuitable when pp is not small, using evidence from the numbers.

00.050.10.150.20.00080.00680.00390.00250123456789101112xProbability
  • B(20, 0.3)
  • Po(6)
B(20, 0.3) and Po(6): the Poisson is more spread out.

Why is Po(6)\text{Po}(6) a poor approximation to B(20,0.3)B(20,0.3)?

Try an exam question

The random variable X∼B(200,0.015)X\sim B(200,0.015). Use a suitable approximation to find P(X≤2)P(X\leq2), stating why the approximation is appropriate.

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Find E(X)E(X).

1 mark, marked against the mark scheme

Revision notes