Lesson 4 of 7 · 15 min
The sample proportion as a random variable
Calculate the sample proportion , see why it varies from sample to sample, and find its exact distribution for small samples from the binomial distribution.
What you will be able to do
- Calculate a sample proportion .
- Explain why is a random variable while is a fixed parameter.
- List the possible values of and find exact probabilities using .
Suppose 30 % of Queensland adults drink coffee every day — so . A researcher surveys a random sample of 50 adults and finds 17 daily coffee drinkers: a sample proportion of . A second researcher surveys a different 50 and finds 13, or . Neither has done anything wrong. Different random samples give different proportions.
Definition
The sample proportion
If members of a random sample of size have an attribute, the sample proportion is
Before the sample is taken, is a random variable, so is a random variable too. Its value varies from sample to sample. The population proportion is a parameter: one fixed number.
In a random sample of 240 Brisbane commuters, 78 travel to work by bus. Calculate the sample proportion of bus commuters.
Draw one sample at a time, then many. Where do the values of cluster? Now change to 200: what happens to the spread?
Each sample adds one value of to the display. After hundreds of samples, the display shows the sampling distribution of : the pattern of values it takes over all possible samples. Two things stand out. The values cluster around , and the cluster is narrower for larger samples.
Key idea
The exact distribution of p̂
In a random sample from a large population, each member independently has the attribute with probability , so the count is binomial: . Then takes the values
So any probability about is a probability about : translate the proportion into a count by multiplying by .
Strictly, sampling without replacement from a finite population makes the trials slightly dependent. When the population is much bigger than the sample (a common guide is at least 10 times bigger), the binomial model is an excellent approximation, and that is the situation in this topic.
Worked example
The exact distribution of p̂ for a small sample
Suppose 30 % of the trees in a large plantation carry a particular fungus. A forester inspects a random sample of 5 trees and records the sample proportion of infected trees.
(a) List the possible values of .
(b) Find the probability distribution of , correct to four decimal places.
(c) Find .
Step 1: Define the count
Let be the number of infected trees in the sample. The plantation is large and the sample random, so , and .
Each tree is like an independent trial with success probability 0.3.
Step 2: List the possible values
can be , so can be .
Step 3: Calculate each probability
:
0 0.2 0.4 0.6 0.8 1 0.1681 0.3602 0.3087 0.1323 0.0284 0.0024 For example, .
Step 4: Translate the inequality
means .
Step 5: Add the probabilities
Try it yourself first, then check the working one step at a time.
Answer
takes the values 0, 0.2, 0.4, 0.6, 0.8, 1 with probabilities 0.1681, 0.3602, 0.3087, 0.1323, 0.0284, 0.0024, and .
Method
Write for the count and .
Convert each statement about into one about by multiplying by .
Use the binomial formula (or a calculator's binomial PD and CD functions).
Watch out for
Treating the six values of as equally likely.
Adding the rounded probabilities: use exact values and round at the end.
A random sample of 5 people is taken from a large population in which 40 % of people own a cat. Calculate .
Common mistake
Mixing up p and p̂
is the population proportion: fixed, usually unknown, and the thing you want to know. is what you calculate from your sample, and it would change if you sampled again. When you write a probability statement, it is always about (or ), never about .
Tip
Proportions and counts
Inequalities translate the same way: with , "" means "", and "" means "". Take care with strict and non-strict inequalities, because only takes whole-number values.
Summary
is the proportion of the sample with the attribute.
is a random variable — it varies between samples; is a fixed parameter.
For a random sample, , so .
Simulated values of cluster around , more tightly for larger samples.
Finished the lesson?
Try the 2 quick checks above and the lesson completes itself, or mark it complete now.