Lesson 6 of 8 · 20 min
The normal distribution
Recognise contexts that suit a normal model, describe the features of the bell curve and the notation , and find normal probabilities with the 68–95–99.7% rule and with technology.
What you will be able to do
- Identify contexts, such as natural variation and measurement error, that suit a normal model.
- Describe the features of the normal pdf: symmetry, centre , spread , inflection points at .
- Read and write , where the second parameter is the variance.
- Estimate probabilities without technology using symmetry and the 68–95–99.7% rule.
- Calculate normal probabilities with a calculator.
Measure the heights of a thousand Year 12 students, the masses of a thousand Bowen mangoes or the fill of a thousand cans of soft drink, and draw a density histogram. Again and again you get the same shape: a single peak in the middle, symmetric, tailing off smoothly on both sides. The curve that fits this shape is the normal distribution, and it is the most used model in all of statistics.
Why so common? A quantity that is the combined result of many small, independent influences (genes, diet and sleep for height; tiny variations in a machine for fill volume) tends to be normally distributed. The ups and downs mostly cancel, so values pile up near the middle, and extreme values need many influences to push the same way, so they are rare.
Key idea
When is a normal model suitable?
Good candidates: naturally occurring variation in a population (heights, masses, blood pressure), manufacturing variation about a target (lengths, volumes), measurement error, and test scores on a well-designed test.
Poor candidates: strongly skewed quantities (incomes, house prices, waiting times, lifetimes), quantities with a hard boundary close to the mean (the number of days absent, when most students have 0), and small discrete counts.
Which variable is most suitably modelled by a normal distribution?
Heights cm with and . Change : the curve slides without changing shape. Change : it spreads out and flattens, keeping total area 1. The shaded area from to stays close to 0.68 for every choice; move the bounds with and to see it.
Definition
The normal distribution
is normally distributed with mean and variance , written
if its pdf is the bell curve with these features:
- it is symmetric about , so the mean, median and mode are all ;
- its peak is at , with height , so a larger gives a lower, wider curve;
- it has points of inflection at and ;
- it is defined for all real and approaches the -axis as without touching it;
- the total area under it is 1.
The standard normal distribution is : mean 0 and standard deviation 1.
The formula for the normal pdf
You need to recognise its graph and features, not memorise or integrate it: no antiderivative can be written with the functions you know, which is why normal probabilities come from technology. Differentiating twice shows at , which is where the inflection points come from.
The mass grams of a bag of rice satisfies . What is the standard deviation of , in grams?
Because every normal curve is the same shape stretched and shifted, the area within a given number of standard deviations of the mean is the same for every normal distribution. Three of these areas are worth knowing by heart.
The 68–95–99.7% rule. About 68% of values lie within of the mean, 95% within and 99.7% within . By symmetry each half is split equally.
| Region | Probability |
|---|---|
| 0.68 | |
| 0.95 | |
| 0.997 | |
| (or ) | 0.5 (exact) |
Worked example
Heights with the 68–95–99.7% rule
The heights cm of Year 12 students at a large school are modelled by . Without technology:
(a) estimate the probability that a student is taller than 186 cm;
(b) estimate ;
(c) estimate how many of the 400 Year 12 students are taller than 186 cm.
Step 1: Write each value in standard deviations from the mean
and , so , and .
The rule only works for whole numbers of standard deviations, so this is always the first step.
Step 2: Upper tail beyond 2σ for (a)
About 95% lie within , leaving 5% split equally between the tails: .
Step 3: Split the interval at the mean for (b)
and , so
The interval is not symmetric about the mean, so treat each side separately.
Step 4: Expected number for (c)
students.
Try it yourself first, then check the working one step at a time.
Answer
(a) (b) (c) about 10 students
Method
Express each bound as .
Split at the mean and use halves of 0.68, 0.95, 0.997.
Expected number = total × probability.
Watch out for
Using 0.05 instead of 0.025 for one tail.
Adding 0.68 and 0.95 for an interval from to : use the halves.
The volume mL of drink in a can is . Without a calculator, estimate as a decimal.
Without technology
Use symmetry (), the 68–95–99.7% rule, and any probabilities the question gives you. Works when the values are a whole number of standard deviations from . Distribution tables are not used.
With technology
Use the normal cdf on your calculator with the lower bound, upper bound, and . For a one-sided tail use a very large bound, such as or . Gives accurate answers for any values.
Worked example
Apple masses with technology
The mass g of an apple from an orchard near Stanthorpe is normally distributed with mean 150 g and standard deviation 12 g.
(a) Find the probability that an apple weighs more than 165 g.
(b) Find the probability that an apple weighs between 140 g and 160 g.
Give answers correct to 4 decimal places.
Step 1: State the distribution
, so and .
Step 2: Sketch and enter the upper tail for (a)
: normal cdf with lower bound 165, upper bound , , , gives .
A sketch shows 165 is above the mean, so the answer must be less than 0.5, a quick check on the tail.
Step 3: Enter both bounds for (b)
: lower bound 140, upper bound 160, , , gives .
Step 4: Check with the rule
140 and 160 are standard deviations either side of the mean, a little less than , so the answer should be a little under 0.68. It is.
Try it yourself first, then check the working one step at a time.
Answer
(a) (b)
Method
Define and state .
Sketch, shade, write the probability statement.
Normal cdf with bounds, and ; use for a one-sided tail.
Watch out for
Entering the variance 144 as .
Test scores satisfy . Use technology to find , correct to 4 decimal places.
Common mistake
Entering the variance as σ
and are the same distribution. Calculators ask for , so enter 8. Entering 64 gives a much flatter curve and a wrong probability.
Exam tip
Define, sketch, state, calculate
For full marks: define the variable and state its distribution (""), sketch a bell curve with the region shaded, write the probability statement , then give the value. Calculator syntax such as "normalcdf(−E99, 75, 65, 8)" on its own is not mathematical communication.
Summary
Normal models suit symmetric, bell-shaped variation from many small independent influences.
: the second parameter is the variance. The standard normal is .
The curve is symmetric about (mean = median = mode), has inflection points at , and has total area 1.
Without technology: , and about 68%, 95% and 99.7% lie within 1, 2 and 3 standard deviations.
With technology: the normal cdf gives from , , and .
Finished the lesson?
Try the 4 quick checks above and the lesson completes itself, or mark it complete now.