Cambridge O Level Statistics
25 notas de revisão de Cambridge International (CAIE) O-Level Statistics, criadas por cookie, em 12 seções que seguem os tópicos do curso.
Notes for Cambridge O Level Statistics (4040), in folders for the twelve topic areas in the specification's order, with one note per sub-topic. Covers the specification for exams from 2025 to 2027, with worked examples and LaTeX working. Delete any folder your course leaves out once the notes are yours.
O que contém
A primeira anotação
Data and its collection / Sampling and surveys
Statistics starts with a group of people or things that we want to know about. The whole group is the population. Collecting data from every member is a census. Collecting data from a smaller part of the population, and using it to say something about the whole, is sampling, and the smaller part is the sample.
A census gives exact figures, but for a large population it is slow and expensive, and sometimes it cannot be done at all. A sample is quicker and cheaper, but it only estimates the population, so the estimate can be wrong. The question in choosing a sample is how to make that estimate as reliable as possible. A sample is representative when it has the same important features as the population (the same mix of ages, sexes, year groups or regions, say), so that what is true of the sample is likely to be true of the population.
General ideas of sampling
Simple random sampling
In simple random sampling every member of the population has the same chance of being chosen, and each selection is independent of the others. To carry it out, every member needs a number, and the sample is found by choosing numbers at random.
Take a school of 400 students. Number them 001 to 400, and choose 40 students by picking random three-digit numbers, such as those from a random number table or a calculator's random function. Ignore any number outside 001 to 400, and ignore a number that has already been chosen. Each student then has the same chance, 40040=101, of being in the sample.
A simple random sample needs a complete list of the population, called a sampling frame. If the frame leaves some people out, the sample cannot represent them.
Systematic sampling
In systematic sampling the members are listed in some order and every kth member is chosen. The interval is found from
and the first member is chosen at random from the first k. For the school above, k=40400=10. If the random start is 7, the sample is students 7, 17, 27, and so on, up to 397, which makes 40 students.
Systematic sampling is quick and spreads the sample evenly along the list. Its weakness is a pattern in the list that happens to repeat with the same period as k. If every tenth student on a register is a boarder, say, a systematic sample could be all boarders or none.
Stratified sampling
Stratified sampling splits the population into groups, called strata, that differ in a way that matters, and then takes a random sample from each stratum in proportion to its size. The number taken from stratum i is
where Ni is the size of the stratum, N the population, and n the sample size.
Suppose the school's 400 students are split by year: 120 in Year 7, 100 in Year 8, 90 in Year 9 and 90 in Year 10. A sample of 40 needs
which adds to 40. Within each year, the students are then chosen at random. Stratified sampling guarantees that every year is represented in proportion, which a simple random sample does not guarantee. It needs the size of each stratum to be known in advance.
Quota sampling
In quota sampling an interviewer is given a set of groups to fill, with a number needed in each, and chooses whoever is convenient until each quota is full. An interviewer asked for 20 men and 20 women over 18 might stand in a shopping centre and stop people as they pass. Quota sampling is cheap and quick, but the interviewer chooses who is included, so the sample is not random, and the people who are easiest to approach may hold different views from the rest. Quotas control the mix of the sample on the characteristics chosen, and nothing else.
Choosing a method
| Method | Needs | Main risk |
|---|---|---|
| Simple random | A complete list of the population | Small groups may be missed by chance |
| Systematic | An ordered list | A pattern in the list that matches the interval |
| Stratified | Known sizes of the strata | Strata chosen badly, so they do not matter to the question |
| Quota | Known proportions in each group | The interviewer's choice of who to approach |
Bias: how it arises and is avoided
A sample is biased when it tends to differ from the population in one direction, so that repeating the survey would keep giving answers that are too high or too low. Bias is a feature of the method, and a larger sample does not remove it. The usual sources are these.
A sampling frame that leaves out part of the population. A telephone directory, for example, leaves out people who have no landline, and tends to leave out younger people.
Non-response, where some of the people chosen do not reply. If the people who do not reply differ from those who do, the results are biased. Following up the non-respondents, or comparing their characteristics with the rest of the sample, reduces this.
Self-selection, where people decide whether to take part. Volunteers for a survey about exercise are often people who already exercise.
Leading questions and interviewer effects, where the wording or the manner of asking pushes people towards a particular answer.
Bias is avoided by choosing the sample at random from a good frame, following up non-respondents, and asking questions neutrally. Testing a questionnaire on a few people first, called a pilot, shows where the wording is unclear.
General ideas of surveys
A survey collects data from people, usually with a questionnaire, an interview or a set of observations. The questions decide what the data can show, so their design matters as much as the sample.
A closed question offers a fixed set of answers, such as tick boxes or a scale. These are quick to answer and easy to count, but they limit what people can say, and the choice of answers can influence the result. An open question lets people answer in their own words. These give richer information but take longer to answer and harder to analyse, because each answer has to be read and grouped.
For closed questions the answer ranges must not overlap. The ranges 0 to 5 and 5 to 10 both include 5, so a respondent cannot tell which box to tick. The ranges 0 to 4, 5 to 9 and 10 or more are clear. Every answer should fit one box, and the boxes should cover every possible answer, with an "other" box where one is needed.
A question should ask one thing at a time. "Do you enjoy sport and music?" cannot be answered yes for one and no for the other, and the answer does not say which was meant.
E mais 24: abra qualquer um deles na lista acima.
Avaliações
Ainda não há avaliações escritas. Adicione estas notas às suas e você poderá ser a primeira pessoa a deixar uma avaliação.