Lesson 7 of 7 · 18 min
Simulation, standardising and when the approximation works
Use repeated random sampling for different p and n to examine the distribution of p̂, standardise p̂ to a z-score, and explain why the normal approximation depends on both n and p.
What you will be able to do
- Standardise a sample proportion using and interpret the result.
- Explain why is also approximately standard normal for large samples.
- Use simulation results to judge how closely the normal approximation fits, for different and .
The normal approximation is an approximation. How good is it? Mathematics gives a partial answer, but the clearest answer comes from simulation: take thousands of random samples, look at the distribution of , and compare it with the normal curve. That is what statisticians did to arrive at the rules of thumb.
Key idea
Standardising p̂
Subtracting the mean and dividing by the standard deviation gives the standardised sample proportion
which has mean 0 and standard deviation 1, and is approximately for large samples. A -score tells you how many standard deviations a sample proportion is from : values beyond about are unusual, and beyond are very rare.
For random samples of 100 from a population with , calculate the -score of a sample proportion of .
In practice you don't know — finding it is the whole point of sampling. So you can't calculate . The natural move is to use the sample's own proportion in its place:
The syllabus asks you to examine, by simulation, whether this is still approximately standard normal. If it is, then about 95 % of samples will have this quantity between and — the fact the next topic turns into confidence intervals.
| Mean of | SD of | Theory | Proportion within | |||
|---|---|---|---|---|---|---|
| 0.5 | 20 | 10 | 0.506 | 0.109 | 0.112 | 0.970 |
| 0.5 | 100 | 50 | 0.500 | 0.050 | 0.050 | 0.940 |
| 0.1 | 20 | 2 | 0.102 | 0.069 | 0.067 | 0.867 |
| 0.1 | 100 | 10 | 0.100 | 0.031 | 0.030 | 0.916 |
| 0.1 | 500 | 50 | 0.100 | 0.013 | 0.013 | 0.951 |
| 0.02 | 100 | 2 | 0.021 | 0.014 | 0.014 | 0.878 |
Read the table in three ways.
- The mean and SD formulas always hold. In every row, the simulated mean is close to and the simulated SD is close to , whatever and are.
- Normality needs a large . For , the proportion within climbs from 0.867 to 0.916 to 0.951 as grows from 20 to 500.
- It also needs away from 0 and 1. With , gives 0.940 but gives only 0.878. What matters is and , the expected numbers with and without the attribute.
1000 simulated values of for , . Here : the distribution piles up against 0 and has a long right tail, so a symmetric normal curve fits it poorly.
For which of these will the distribution of be closest to normal?
Start at , (). Increase until the normal curve fits well. What is when it starts to look right? Try the same with .
Worked example
Standardise with p and with p̂
Suppose 40 % of Australian adults have a streaming music subscription. In a random sample of 150 adults, 69 have one.
(a) Calculate the standardised value .
(b) Calculate , correct to two decimal places, and compare it with (a).
(c) Is this sample unusual?
Step 1: Find the sample proportion
Step 2: Standardise using p
, so .
Step 3: Standardise using p̂
, so .
When is unknown, is the best available estimate for it in the standard deviation.
Step 4: Compare
The two values are close, because is close to , so the two standard deviations are almost the same. For large samples this is typical, which is why the second form is still approximately .
Step 5: Judge
Both values are well within , so a sample proportion of 0.46 is not unusual for samples of 150 when .
Try it yourself first, then check the working one step at a time.
Answer
(a) . (b) , very close to (a). (c) No: the sample is less than 2 standard deviations from .
Method
Calculate , then the standard deviation, then .
With unknown, use instead.
is unusual; is very unusual.
Watch out for
Dividing by the variance instead of the standard deviation.
Mixing the two forms, e.g. .
Common mistake
When the normal model breaks
With small, the normal model gives impossible answers: a positive probability that , and a symmetric spread when the real distribution is piled up against 0. If the standardised quantity can't even be calculated (you would divide by 0). In these cases use the exact binomial distribution instead.
Worked example
When the normal approximation fails
A factory says 3 % of its light globes are faulty. An inspector tests a random sample of 50 globes.
(a) Find the mean and standard deviation of .
(b) A student uses the normal approximation to calculate . Find the value they get, correct to three decimal places, and explain why it shows the approximation is poor here.
(c) Use the binomial distribution to find , and describe the shape of the true distribution of .
Step 1: Mean and standard deviation
and .
Step 2: Apply the normal model
Step 3: Explain why this is a problem
A proportion can't be negative, so the true probability is 0. The normal model puts about 11 % of its probability on impossible values. The rule of thumb predicts this: , far below 10.
Step 4: Use the exact distribution
and .
Step 5: Describe the shape
About 22 % of samples have no faulty globes at all, and the distribution of is piled up at with a long right tail. It is strongly skewed to the right, not symmetric and bell-shaped.
Try it yourself first, then check the working one step at a time.
Answer
(a) , . (b) The normal model gives , but negative proportions are impossible, so the model is poor; is far too small. (c) ; the true distribution is strongly right-skewed.
Method
Check and before trusting the normal model.
A positive probability for an impossible value is a sign that the model doesn't fit.
When is small, use the binomial distribution.
Watch out for
Reporting 0.107 as the probability of a negative proportion without questioning it.
Thinking the mean and SD formulas fail too: they are exact. Only normality fails.
Exam tip
Explaining the approximation
If asked to explain or justify whether the normal approximation is appropriate, calculate and , compare them with the rule of thumb, and say what that means for the shape (symmetric and bell-shaped, or skewed). Refer to simulation or the exact binomial when the question gives you that evidence.
Summary
is approximately for large samples; is unusual.
With in place of the unknown , is also approximately standard normal for large samples.
and hold for every and ; normality needs large and .
Simulation shows the approximation improving as grows, and worsening as moves towards 0 or 1.
When or is small, use the exact binomial distribution.
Finished the lesson?
Try the 2 quick checks above and the lesson completes itself, or mark it complete now.