In statistics, a sample is a subset from a larger population.
Key words
sample
a chosen set from a larger group
unbiased
no outcome favoured by selection
empirical
based on observation
theoretical probability distribution
exact chance of each result given by maths
How to do it
Pick an unbiased sample from the population.
Record each outcome of the experiment.
Count how many times each result occurs.
Compare observed proportions with theoretical probabilities; larger samples give closer match.
Watch out for
Increasing the sample makes the observed frequency closer to the true chance.
Remember: A larger sample means better agreement with theory.
Sample size and probability Core
In statistics we compare what we actually see – the empirical distribution – with what theory predicts – the theoretical distribution. As the sample becomes larger, random error shrinks and the empirical pattern moves closer to the theoretical one. This is known as the law of large numbers.
Key words
sample
A set of observations taken from the population.
population
The whole group of interest; every member that could be measured.
unbiased sample
Every member has an equal chance of being chosen (simple random sampling).
theoretical probability distribution
Mathematical model giving exact probabilities, e.g. binomial or normal.
How to do it
Identify the population and whether your sample is unbiased.
State the theoretical probability distribution that applies to the situation and note its key parameters (for a binomial experiment: number of trials n and success probability p).
From the data given, calculate the empirical proportion or frequency of the outcome of interest.
Use the law of large numbers: as n increases the empirical proportion tends towards the theoretical p. Predict how the proportion will change when the sample size grows.
When a numerical estimate is needed, compute the expected number of successes as n\(\times\) p and then divide by the total number of trials to obtain the expected empirical proportion.
Watch out for
A small sample may give a misleading proportion – e.g., 22/40 = 0.55 is below the true 0.6 because of random error. With more flips the observed proportion will fluctuate less and converge to the theoretical value.
Bias would shift the average away from the true probability; it is not reduced by larger samples.
Remember: When designing an experiment first decide on an acceptable margin of error; this tells you roughly how many observations are required to achieve the desired level of confidence.
Sample size and probability Stretch
To check how a real sample compares with theory you calculate the experimental proportion of an event, then see if it is close to the theoretical probability. As the sample size grows this closeness improves – this is the law of large numbers.
Key words
unbiased
no systematic error in selection
experimental proportion
number of observed successes ÷ total trials
How to do it
Write down the theoretical probability of the event from the given information.
Count how many times the event occurs in the sample and calculate the experimental proportion \(\frac{c}{n}\).
Compare this proportion with the theoretical probability; note the difference or ratio.
Remember that with a larger \(n\) the experimental proportion will usually move closer to the theoretical value.
If the sample is very large, you can say it 'tends toward' the theory – a basic form of the law of large numbers.
Watch out for
Assume a small sample is 'exactly' theoretical – errors are larger when \(n\) is small.
Mix up percentages and fractions; always convert before comparing.
Ignore that sampling without replacement slightly changes probabilities if the population is small.
Remember: This concept links to binomial probability (probability of a given number of successes in many trials) and the rule of 1\% in data analysis. It underpins the idea that repeated experiments will average out random variation.
✏️ Try 3 quick questions
The theoretical probability of getting heads on a fair coin is 0.5. If you flip it 8 times, how many heads must be observed to have an empirical probability of exactly 0.5?
A fair die was rolled 40 times and the observed proportion of sixes is \(\frac{1}{4}\). How many sixes were rolled?
A gardener plants 77 seedling trees; each seedling independently has a probability 0.13 of producing fruit during the first year. What is the expected number of fruit‑bearing trees? Give your answer to two decimal places.
There are more on this in 4 Triangles, with hints and worked answers, and real past-paper questions. Carry on there.
Practise it with real exam questions
4 Triangles has quick questions on sample size and probability, real GCSE past-paper questions with their mark schemes, and a revision plan that picks what to work on next.
Questions on this topic for a lesson, homework or revision - a quick "do now", misconception questions, small steps that change one thing at a time, and practice, with answers on the last page.