Normal approximation to the binomial and Poisson

Normal approximation to the binomial and Poisson

Spec 3.2 · Edexcel International A Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • Use B(n,p)≈N(np,npq)\mathrm{B}(n,p)\approx\mathrm{N}(np,npq) when nn is large and pp is close to 0.50.5 (np>5np>5, nq>5nq>5).
  • Use Po(λ)≈N(λ,λ)\mathrm{Po}(\lambda)\approx\mathrm{N}(\lambda,\lambda) when λ\lambda is large (usually >10>10).
  • The second parameter of N(μ,σ2)\mathrm{N}(\mu,\sigma^2) is the variance.
  • Always apply a continuity correction: move the boundary 0.50.5 outwards if the integer is included.

When it is appropriate

When the discrete distribution is large and roughly symmetrical.

Binomial and Poisson probabilities are tedious to add up when nn or λ\lambda is large. A normal approximation replaces the discrete distribution with a continuous one with the same mean and variance.

If pp is very small or very large, the distribution is skewed and the approximation is poor.

24681012141618200.020.040.060.080.10.120.140.160.180.2xymean 10normal curve
Illustrative: a large, roughly symmetrical distribution has the bell shape of a normal curve (here N(10, 5))

Binomial B(n, p)

Good when:
n large, p close to 0.5
Usual check:
np > 5 and nq > 5

Poisson Po(λ)

Good when:
λ large
Usual check:
λ > 10

Which distribution is best approximated by a normal distribution?

Approximating the binomial

X≈N(np, npq)X\approx\mathrm{N}(np,\,npq), where q=1−pq=1-p.

If X∼B(n,p)X\sim\mathrm{B}(n,p) with nn large and pp near 0.50.5, the mean is npnp and the variance is npqnpq. The second parameter of N(μ,σ2)\mathrm{N}(\mu,\sigma^2) is the variance, so standardise with the standard deviation σ=npq\sigma=\sqrt{npq}.

For X∼B(80,0.4)X\sim\mathrm{B}(80,0.4): μ=32\mu=32, σ2=19.2\sigma^2=19.2, so X≈N(32,19.2)X\approx\mathrm{N}(32,19.2).

24681012141618200.020.040.060.080.10.120.140.160.180.2xymean np = 10N(10, 7.5)
X ~ B(40, 0.25) is approximated by N(np, npq) = N(10, 7.5)
  • B(n, p)≈N(np, npq)\approx\mathrm{N}(np,\ npq)

Worked example

X∼B(80,0.4)X\sim\mathrm{B}(80,0.4). Write down the normal approximation.

X∼B(200,0.3)X\sim\mathrm{B}(200,0.3). Which is the normal approximation?

Approximating the Poisson

X≈N(λ,λ)X\approx\mathrm{N}(\lambda,\lambda), since the mean and variance are both λ\lambda.

If X∼Po(λ)X\sim\mathrm{Po}(\lambda) with λ\lambda large, the Poisson distribution has mean and variance both equal to λ\lambda, so the standard deviation is λ\sqrt\lambda.

For X∼Po(25)X\sim\mathrm{Po}(25): X≈N(25,25)X\approx\mathrm{N}(25,25) with σ=5\sigma=5.

510152025300.020.040.060.080.10.12xymean λ = 16N(16, 16)
X ~ Po(16) is approximated by N(16, 16): the mean and variance are both λ
  • Po(λ)≈N(λ, λ)\approx\mathrm{N}(\lambda,\ \lambda)

X∼Po(49)X\sim\mathrm{Po}(49). What is the normal approximation?

The continuity correction

XX takes whole numbers but the normal variable is continuous, so each integer kk becomes the block k−0.5<Y<k+0.5k-0.5<Y<k+0.5.

The adjustment is the continuity correction. Rewrite strict inequalities as inclusive ones first: X<20X<20 means X≤19X\le19, so the boundary is 19.519.5.

Ask whether the integer is included. If it is, move the boundary outwards by 0.50.5. If it is not, move it inwards.

2727.52828.52929.53030.53131.53232.5333030.5

P(X≤k)P(X\le k)

  • P(Y<k+0.5)P(Y<k+0.5)

P(X<k)P(X<k)

  • P(Y<k−0.5)P(Y<k-0.5)

P(X≥k)P(X\ge k)

  • P(Y>k−0.5)P(Y>k-0.5)

P(X>k)P(X>k)

  • P(Y>k+0.5)P(Y>k+0.5)

Write P(X>12)P(X>12) with a continuity correction.

Worked examples

Approximate, apply the correction, standardise, then read the table.

State the approximating distribution. Apply the continuity correction. Standardise with z=y−μσz=\frac{y-\mu}{\sigma} and use Φ\Phi.

For an inverse problem, solve k+0.5−μσ≥z\frac{k+0.5-\mu}{\sigma}\ge z and round up to an integer.

101520253035400.020.040.060.080.1xy29.5N(25, 25)
Po(25)≈N(25,25)\mathrm{Po}(25)\approx\mathrm{N}(25,25): P(X≥30)≈P(Y>29.5)P(X\ge30)\approx P(Y>29.5) is the shaded area.

Worked example

X∼Po(25)X\sim\mathrm{Po}(25). Find P(X≥30)P(X\ge30).

Worked example

X∼B(150,0.45)X\sim\mathrm{B}(150,0.45). Find P(60≤X<75)P(60\le X<75).

Worked example

X∼Po(36)X\sim\mathrm{Po}(36). Find the smallest kk with P(X≤k)≥0.95P(X\le k)\ge0.95.

X∼Po(16)X\sim\mathrm{Po}(16). Find the zz value used to approximate P(X<20)P(X<20).

Try an exam question

X∼B(120,0.4)X\sim\mathrm{B}(120,0.4). Use a normal approximation to find P(X≥55)P(X\ge55).

[5 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Which normal distribution approximates XX?

1 mark, marked against the mark scheme

Revision notes