Sampling and a large data set
Introduzido| English | Significado |
|---|---|
| sampling frame/ˈsæmplɪŋ freɪm/ | quadro de amostragem |
A poll of 2.4 million that got it wrong, 1936
- In 1936, the magazine Literary Digest mailed 10 million ballots and received 2.4 million replies. It predicted that Alf Landon would beat President Franklin Roosevelt. Roosevelt won 46 of 48 states. The names had come from phone books, car registrations and the magazine's own readers — wealthier people in the Great Depression. A sample is only as good as the list it is drawn from: the sampling frame 抽样框.
- A small, well-chosen sample beats a huge biased one; George Gallup predicted Roosevelt correctly with about 50,000 people.
Population, sample, frame
- The population is everyone you want to know about; a sample is the part you measure.
- The sampling frame is the list you actually choose from — if it leaves people out, the sample is biased.
A survey of a town's adults uses the list of members of one gym. What is the main problem?
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Non-members are left out.
Sampling methods
- Simple random: every sample of size $n$ is equally likely (number the frame, use random numbers). Systematic: every $k$th item from a random start.
- Stratified: sample each group in proportion to its size. Quota and opportunity samples are quick but not random.
Sampling and a large data set
Identify the population, sampling unit and frame
Allocate proportionally, then explain why random selection within each group still matters.
Match each description to the sampling method.
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
Definitions of the methods.
A systematic sample of 40 is taken from a list of 1000. What is the gap $k$?
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
$1000 \div 40$.
Worked example: a stratified sample
A school has 90 Year 12 and 60 Year 13 students. Choose a stratified sample of 30.
- Fraction: $\tfrac{30}{150} = \tfrac{1}{5}$ of each year group.
- Allocate: Year 12: $\tfrac{1}{5} \times 90 = 18$; Year 13: $\tfrac{1}{5} \times 60 = 12$.
- Choose: number each year group's list and pick 18 and 12 students at random.
A firm has 120 office staff and 80 factory staff. How many office staff are in a stratified sample of 25?
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
$\tfrac{120}{200} \times 25$.
The large data set
- AQA's large data set is real data you study in advance: questions expect you to know its variables, units and limits.
- Clean it before use: remove impossible or missing values, and say why.
A large data set records a daily rainfall of $-3$ mm. What should you do?
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Rainfall cannot be negative.
Judging a sample
- Ask: does the frame cover the population? Is the method random? Is the sample big enough?
- A systematic sample can be biased if the list has a pattern every $k$th item.
A larger sample does not fix a biased frame. The Literary Digest had 2.4 million replies and was still wrong, because the list left out most poorer voters. Fix the frame first, then the size.
A sample of 2 million is always more reliable than a random sample of 2000.
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
A biased frame stays biased however large.
Define the population, check that the sampling frame covers it, and choose a method: random, systematic, stratified, quota or opportunity. Allocate stratified samples in proportion to group sizes. Bias comes from the frame and the method, not just from the sample size.
A student wants to find how long sixth-formers spend on homework. She asks 20 friends at lunch. Name the method, give two weaknesses, and describe a better sample of 20 from 200 students (120 Year 12, 80 Year 13).
Evaluating and improving a sample.