4.9 · SLPaper 1Paper 2

Normal distribution

In a minute

Heights, masses and many other measurements are often modelled by the : a symmetric bell-shaped curve centred on the . Its probabilities are areas under the curve. You estimate them with the , find them with your GDC, and run the GDC backwards to find a value from a probability.

You'll be able to

  • Describe the normal curve, its properties and where it occurs, read X∼N(μ,σ2)X \sim \mathrm{N}(\mu, \sigma^{2}), and sketch it with a region shaded.
  • Use the 68–95–99.7 rule to estimate probabilities and expected numbers, and to judge whether a model is sensible.
  • Find normal probabilities with your GDC and say what they mean in context.
  • Find a value from a probability with , for a lower tail, an upper tail or a middle region.
Before you start3 quick questions on what this builds on
Check yourself
A data set has variance 6464. Find its .
Check yourself
The probability that a seed fails to sprout is 0.150.15. How many of 240240 seeds do you expect to fail?
Check yourself
A XX has P(X≤3)=0.71\mathrm{P}(X \le 3) = 0.71. Find P(X>3)\mathrm{P}(X > 3).

Revise first: 4.2 Data presentation, 4.3 Central tendency and dispersion, 4.5 Probability foundations and expected value and 4.7 Discrete random variables and expected value

The normal curve

Draw a of 2000 adult heights with narrow bars. Most heights sit near the average, and extreme ones are rare on both sides. The outline is a smooth, symmetric bell: the normal curve.

This shape appears when many small, independent effects add up. Heights, masses, machine-made lengths and measurement errors are often close to normal. Incomes and waiting times are usually skewed, so they are not.

A height is a : it can take any value in a range, unlike the separate values in 4.7. Its probabilities are areas under a curve.

The second number in N(μ,σ2)\mathrm{N}(\mu, \sigma^{2}) is the variance, not the standard deviation. So X∼N(30,16)X \sim \mathrm{N}(30, 16) has σ=4\sigma = 4. Questions often write N(30,42)\mathrm{N}(30, 4^{2}) to make σ\sigma easy to see.

Bell-shaped normal curve, symmetric about a dashed vertical line at the mean mu. The area under the curve between two values a and b to the right of the mean is shaded and labelled as the probability that X lies between a and b. The whole area under the curve is 1. μ a b shaded area = P(a < X < b) total area = 1 mean = median = mode
The probability that XX lies between aa and bb is the area under the curve between them.

1. The curve is symmetric about the vertical line through μ\mu. So the mean, median and mode all equal μ\mu.

2. The total area under the curve is 11, because XX is certain to take some value.

3. The curve gets closer and closer to the horizontal axis on both sides but never touches it. Values far from μ\mu are possible, just very unlikely.

4. μ\mu sets where the centre is. σ\sigma sets the spread: a larger σ\sigma gives a lower, wider curve, because the area must stay 11.

Two normal curves with the same mean mu. The curve with the smaller standard deviation is tall and narrow. The curve with the larger standard deviation is lower and wider. Both enclose a total area of 1. μ smaller σ: taller, narrower larger σ: lower, wider each total area = 1
Same μ\mu, different σ\sigma: the wider curve is lower, so both areas are 11.

Symmetry gives some probabilities with no working: half the area lies on each side of μ\mu, and tails the same distance dd from μ\mu have equal areas.

Learn this
X∼N(μ,σ2):P(X<μ)=P(X>μ)=0.5,P(X<μ−d)=P(X>μ+d)X \sim \mathrm{N}(\mu, \sigma^{2}): \quad \mathrm{P}(X < \mu) = \mathrm{P}(X > \mu) = 0.5, \qquad \mathrm{P}(X < \mu - d) = \mathrm{P}(X > \mu + d)

To sketch a normal curve, draw a symmetric bell that flattens towards the axis without touching it. Mark μ\mu in the centre, label each bound on the correct side of μ\mu, and shade the region you want.

Worked exampleThe hand spans SS cm of adults in a town are modelled by S∼N(20.4,2.25)S \sim \mathrm{N}(20.4, 2.25). Write down (a) the standard deviation of SS, (b) P(S<20.4)\mathrm{P}(S < 20.4).

Try it first, then reveal one step at a time.

  1. (a) Read the model: the mean is 20.420.4 and the variance is 2.252.25. The standard deviation is the square root of the variance: 2.25=1.5\sqrt{2.25} = 1.5 cm.
Worked exampleThe masses RR grams of bags of rice are normally distributed, with P(R<994)=P(R>1010)=0.07\mathrm{P}(R < 994) = \mathrm{P}(R > 1010) = 0.07. Find the mean mass, mm grams, and p=P(994<R<m)p = \mathrm{P}(994 < R < m).

Try it first, then reveal one step at a time.

  1. The two tails have equal areas, so 994994 and 10101010 are the same distance from the mean.
Check yourself
Which of these is most likely to be well modelled by a normal distribution?
Check yourself
The times TT seconds that club runners take for 400400 m are modelled by T∼N(52,42)T \sim \mathrm{N}(52, 4^{2}). Find P(T=52)\mathrm{P}(T = 52).
Check yourself
X∼N(50,62)X \sim \mathrm{N}(50, 6^{2}) and P(X<k)=0.8\mathrm{P}(X < k) = 0.8. Where is kk?

The 68–95–99.7 rule

Every normal curve has the same shape: μ\mu slides it along and σ\sigma stretches it. So the fraction within one standard deviation of the mean is the same for every normal distribution.

Learn this
X∼N(μ,σ2):P(μ−σ<X<μ+σ)≈0.68P(μ−2σ<X<μ+2σ)≈0.95P(μ−3σ<X<μ+3σ)≈0.997X \sim \mathrm{N}(\mu, \sigma^{2}): \quad \begin{aligned} \mathrm{P}(\mu - \sigma < X < \mu + \sigma) &\approx 0.68 \\ \mathrm{P}(\mu - 2\sigma < X < \mu + 2\sigma) &\approx 0.95 \\ \mathrm{P}(\mu - 3\sigma < X < \mu + 3\sigma) &\approx 0.997 \end{aligned}

This is the , for normal distributions only. Learn it: it is not in the formula booklet. The values are rounded; the GDC gives 0.68270.6827 for one standard deviation.

Normal curve with vertical lines at one, two and three standard deviations either side of the mean. Brackets underneath show that about 68 percent of values lie within one standard deviation of the mean, about 95 percent within two, and about 99.7 percent within three. The two strips next to the mean each hold about 34 percent. μ − 3σ μ − 2σ μ − σ μ μ + σ μ + 2σ μ + 3σ 34% 34% 13.5% 13.5% about 68% about 95% about 99.7%
Split each band in half at μ\mu: the curve is symmetric.

Use symmetry to split the bands. Half of 68%68\% is 34%34\%, between μ\mu and μ+σ\mu + \sigma. Half of 95%95\% is 47.5%47.5\%, so 47.5%−34%=13.5%47.5\% - 34\% = 13.5\% lies between μ+σ\mu + \sigma and μ+2σ\mu + 2\sigma.

Half of 99.7%99.7\% is 49.85%49.85\%. So 0.15%0.15\% lies above μ+3σ\mu + 3\sigma, and the same below μ−3σ\mu - 3\sigma.

Worked exampleThe lengths LL cm of adult salmon in a river are modelled by L∼N(72,52)L \sim \mathrm{N}(72, 5^{2}). Use the 68–95–99.7 rule to estimate P(62<L<82)\mathrm{P}(62 < L < 82).

Try it first, then reveal one step at a time.

  1. Write each bound as standard deviations from the mean: 62=72−2×5=μ−2σ62 = 72 - 2 \times 5 = \mu - 2\sigma and 82=72+2×5=μ+2σ82 = 72 + 2 \times 5 = \mu + 2\sigma.
Worked exampleThe battery life BB hours of a phone model is modelled by B∼N(18,1.52)B \sim \mathrm{N}(18, 1.5^{2}). A shop tests 20002000 phones. Estimate how many have a battery life between 19.519.5 and 22.522.5 hours.

Try it first, then reveal one step at a time.

  1. Convert the bounds: 19.5=18+1.5=μ+σ19.5 = 18 + 1.5 = \mu + \sigma and 22.5=18+3×1.5=μ+3σ22.5 = 18 + 3 \times 1.5 = \mu + 3\sigma.
Check yourself
Commute times in a city are normally distributed. About 99.7%99.7\% of them lie between 2626 and 5050 minutes, an interval centred on the mean. Estimate the standard deviation, in minutes.
Check yourself
The heights HH cm of sunflowers in a field are modelled by H∼N(180,152)H \sim \mathrm{N}(180, 15^{2}). Use the 68–95–99.7 rule to estimate P(H>195)\mathrm{P}(H > 195).
Check yourself
A test is marked out of 5050. A teacher models the marks by N(40,52)\mathrm{N}(40, 5^{2}). Use the 68–95–99.7 rule to comment on the model.

Normal probabilities on your GDC

The rule only works at whole numbers of standard deviations. For other bounds, use your GDC's normal cdf.

1. Sketch the curve, mark μ\mu and the bounds, and shade the area you want.

2. Write the probability and the model, such as P(248<V<255)\mathrm{P}(248 < V < 255) with V∼N(252,3.22)V \sim \mathrm{N}(252, 3.2^{2}).

3. Enter the lower bound, upper bound, μ\mu and σ\sigma. Enter σ\sigma, not the variance.

4. For less than bb, use the lower bound −1099-10^{99} (type -1E99). For more than aa, use the upper bound 109910^{99} (1E99).

Worked exampleThe volume VV ml of juice in a carton is modelled by V∼N(252,3.22)V \sim \mathrm{N}(252, 3.2^{2}). Use your calculator to find the probability that a carton holds between 248248 and 255255 ml. Give your answer to 3 s.f.

Try it first, then reveal one step at a time.

  1. Sketch: 252252 in the centre, 248248 to its left and 255255 to its right. Shade between them.
Worked exampleThe finishing times TT minutes in a 55 km park run are modelled by T∼N(27.5,3.62)T \sim \mathrm{N}(27.5, 3.6^{2}). (a) Find the probability that a runner finishes in under 2424 minutes. (b) 180180 runners take part. Find the expected number who finish in under 2424 minutes.

Try it first, then reveal one step at a time.

  1. (a) Sketch: 2424 is left of the mean, so expect a small left tail, under 0.50.5.
Worked exampleA café owner models the time TT minutes a customer waits for a coffee by T∼N(4,2.52)T \sim \mathrm{N}(4, 2.5^{2}). Find P(T<0)\mathrm{P}(T < 0), and comment on the model.

Try it first, then reveal one step at a time.

  1. Write P(T<0)\mathrm{P}(T < 0). Normal cdf with lower bound −1099-10^{99}, upper bound 00, μ=4\mu = 4, σ=2.5\sigma = 2.5 gives 0.0547992…0.0547992\ldots
Check yourself
The birth masses MM kg of lambs on a farm are modelled by M∼N(4.6,0.72)M \sim \mathrm{N}(4.6, 0.7^{2}). Find the probability that a lamb weighs more than 5.55.5 kg. Give your answer to 3 s.f.
Check yourself
X∼N(30,16)X \sim \mathrm{N}(30, 16). Find P(X>25)\mathrm{P}(X > 25), to 3 s.f.

Inverse normal: from a probability to a value

Sometimes you know the area and need the boundary: the time the fastest 10%10\% beat, or the length only the longest 5%5\% exceed. This is an calculation.

In inverse normal questions μ\mu and σ\sigma are always given. You only need the area, and which side of kk it is on.

1. P(X<k)=p\mathrm{P}(X < k) = p: enter the area pp with the tail set to left (the default).

2. P(X>k)=p\mathrm{P}(X > k) = p: the area to the left of kk is 1−p1 - p. Enter 1−p1 - p, or enter pp with the tail set to right if your GDC has that option.

Normal curve with mean mu and a value k to the right of the mean. The small tail to the right of k is shaded darker and labelled area p. Everything to the left of k is shaded lighter and labelled area 1 minus p. μ k area p area 1 − p (left of k)
If the area to the right of kk is pp, the area to its left is 1−p1 - p.
Learn this
P(X>k)=p  ⟺  P(X<k)=1−p(X continuous)\mathrm{P}(X > k) = p \iff \mathrm{P}(X < k) = 1 - p \quad (X \text{ continuous})

3. A middle region of probability pp, centred on μ\mu: the two tails share 1−p1 - p, so each holds 1−p2\frac{1 - p}{2}. The lower bound has 1−p2\frac{1 - p}{2} to its left, and the upper bound has 1+p2\frac{1 + p}{2}.

Worked exampleThe time TT minutes to assemble a flat-pack desk is modelled by T∼N(47,6.52)T \sim \mathrm{N}(47, 6.5^{2}). The fastest 10%10\% of assemblers win a badge. Find the time tt they must beat.

Try it first, then reveal one step at a time.

  1. Sketch: the fastest times are a small left tail, so tt is below 4747.
Worked exampleThe lengths CC cm of cucumbers from a greenhouse are modelled by C∼N(31,2.42)C \sim \mathrm{N}(31, 2.4^{2}). The longest 5%5\% are sold as extra long. Use your calculator to find the shortest length kk an extra-long cucumber can have.

Try it first, then reveal one step at a time.

  1. Write the condition: P(C>k)=0.05\mathrm{P}(C > k) = 0.05.
Worked exampleThe wingspans WW cm of a bird species are modelled by W∼N(38.5,1.82)W \sim \mathrm{N}(38.5, 1.8^{2}). The middle 90%90\% of wingspans lie between pp and qq, symmetric about the mean. Find pp and qq.

Try it first, then reveal one step at a time.

  1. Outside the middle 90%90\% lies 10%10\%, shared equally: 5%5\% in each tail.
Check yourself
Monthly mobile data use, DD GB, is modelled by D∼N(6.2,1.42)D \sim \mathrm{N}(6.2, 1.4^{2}). 30%30\% of users use more than kk GB. Find kk.
Check yourself
Weekly screen time, SS hours, for students at a school is modelled by S∼N(21,52)S \sim \mathrm{N}(21, 5^{2}). Find the lower quartile of SS. Give your answer to 3 s.f.

Common mistakes

  • Typing the variance into the GDC's standard deviation field. In N(μ,σ2)\mathrm{N}(\mu, \sigma^{2}) the second number is σ2\sigma^{2}, so take its square root first.
  • Using pp as a left-hand area when P(X>k)=p\mathrm{P}(X > k) = p. Enter 1−p1 - p, or set the tail to right.
  • Putting all the leftover area in one tail. For the middle 80%80\%, each tail holds 10%10\%: the bounds have 0.10.1 and 0.90.9 to their left.
  • Rounding an expected number to a whole number, such as 3030 for 29.829.8, or using a rounded probability to find it. Use the stored value and give 3 s.f.
  • Copying a GDC display such as 6.13E-3 as 6.136.13. A probability can't be more than 11: write 0.006130.00613.
  • Giving a bare answer to 2 s.f., such as 0.170.17 for 0.165…0.165\ldots With no working shown, it loses the accuracy mark and there is no method mark to fall back on. Write the set-up and give 3 s.f.

Exam tips

  • Find / calculate: write the probability statement, such as P(T<24)\mathrm{P}(T < 24), or the GDC syntax with its four numbers, then the answer to 3 s.f. The written line earns the method mark.
  • Sketch: a symmetric bell that flattens towards the axis, the mean in the centre, each bound labelled on the correct side of the mean, and the region shaded.
  • Write down: P(X<μ)=0.5\mathrm{P}(X < \mu) = 0.5 needs no working. Use symmetry and the 68–95–99.7 rule for quick answers and to sense-check GDC results.
  • Comment on a model: find the probability of an impossible value, such as a negative time or a mark above the maximum, and say what it means in context.

Key takeaways

Learn thisX∼N(μ,σ2):P(X<μ)=P(X>μ)=0.5,P(X<μ−d)=P(X>μ+d)X \sim \mathrm{N}(\mu, \sigma^{2}): \quad \mathrm{P}(X < \mu) = \mathrm{P}(X > \mu) = 0.5, \qquad \mathrm{P}(X < \mu - d) = \mathrm{P}(X > \mu + d)
Learn thisX∼N(μ,σ2):P(μ−σ<X<μ+σ)≈0.68P(μ−2σ<X<μ+2σ)≈0.95P(μ−3σ<X<μ+3σ)≈0.997X \sim \mathrm{N}(\mu, \sigma^{2}): \quad \begin{aligned} \mathrm{P}(\mu - \sigma < X < \mu + \sigma) &\approx 0.68 \\ \mathrm{P}(\mu - 2\sigma < X < \mu + 2\sigma) &\approx 0.95 \\ \mathrm{P}(\mu - 3\sigma < X < \mu + 3\sigma) &\approx 0.997 \end{aligned}
Learn thisP(X>k)=p  ⟺  P(X<k)=1−p(X continuous)\mathrm{P}(X > k) = p \iff \mathrm{P}(X < k) = 1 - p \quad (X \text{ continuous})
Learn thisX∼N(μ,σ2)X \sim \mathrm{N}(\mu, \sigma^{2}): a symmetric bell-shaped curve about μ\mu, with total area 11. Mean, median and mode all equal μ\mu. The second number is the variance.
Learn thisFor a continuous random variable, P(X=a)=0\mathrm{P}(X = a) = 0, so << and ≤\le give the same probability.
Learn thisNormal cdf takes lower bound, upper bound, μ\mu and σ\sigma. Use −1099-10^{99} as the lower bound for less than, and 109910^{99} as the upper bound for more than.
Learn thisInverse normal, with the tail set to left, returns kk from P(X<k)=p\mathrm{P}(X < k) = p. For a middle region of probability pp centred on μ\mu, each tail holds 1−p2\frac{1 - p}{2}.
Learn thisExpected number of nn independent items in a region =n×p= n \times p, using the unrounded probability. Give it to 3 s.f.: it need not be a whole number.

Practise this

26 past-paper questions on Normal distribution in the AI HL bank, each with a mark scheme and a worked explanation.

Practise Normal distribution →