Collecting data · GCSE Statistics

Questionnaires

Open questions let people answer in any way. They give detail but are hard to analyse.
Closed questions have a fixed set of answers, such as tick boxes. They are easy to analyse but limit answers.
Good response boxes don't overlap, cover every possible answer, and include a time frame where needed (such as "per week").
Avoid leading questions ("Don't you agree that…?"), vague words ("often"), and unnecessary personal questions.
Questionnaires can be posted, handed out or put online. Response rates may be low, and people who respond may not be typical.

For sensitive questions (such as "Have you ever cheated in a test?"), use the random response technique. Each person flips a coin: heads, they answer "yes"; tails, they answer truthfully. Nobody knows which, so people answer honestly.
Example: 100 people, 60 say yes. About 50 got heads and said yes, so 10 of the 50 answering truthfully said yes. The estimate is 10 ÷ 50 = 20%.

Interviews allow follow-up questions and a high response rate, but take time, and the interviewer may influence answers (interviewer bias).
Observation records what people actually do, and data logging uses machines to record automatically.
Laboratory experiments are carried out in controlled conditions. Field experiments happen in a natural setting with some control. Natural experiments look at changes that happen without the researcher controlling them.
A control group doesn't receive the treatment, so results can be compared.

Before analysing, clean the data: remove errors and impossible values, make units consistent, and deal with missing answers.

Open question: A question that lets people answer in any way.
Closed question: A question with a fixed set of answers.
Leading question: A question that pushes people towards a particular answer.
Random response technique: A method using chance to get honest answers to sensitive questions.
Control group: A group that doesn't receive the treatment, for comparison.
Interviewer bias: When an interviewer influences the answers given.
Data cleaning: Removing errors and making data consistent before analysis.
Field experiment: An experiment in a natural setting with some control.