The research process
Sociologists follow steps much like other scientists: choose a question, review earlier work, state a hypothesis (a testable statement about a relationship between variables), choose a design, collect data, analyse it and report. An independent variable is the presumed cause; the dependent variable is the outcome. A key task is the operational definition, the exact way a vague idea such as "social isolation" will be measured, for example the number of people a person talks with in a week.
Quality of measurement
- Reliability: repeated measurement gives consistent results.
- Validity: the measure captures the concept it claims to. A bathroom scale that always reads 2 kilos too high is reliable but not valid.
Common methods
- Surveys: questionnaires or interviews given to many people. They give broad data, but wording and honest answers can be problems.
- Ethnography and participant observation: the researcher spends long periods inside a group. This gives deep, rich detail, but results may not generalise.
- Experiments: control conditions and assign people to groups. Rarer in sociology because many social conditions cannot be assigned.
- Secondary data analysis: reusing existing data, such as census records. Cheap but not designed for your question.
- Content analysis: coding the themes in documents, media or images.
Quantitative data are numbers; qualitative data are words, observations and meanings. Many studies mix both.
Sampling
The population is everyone the study is about. A sample is the subset actually studied. In a random sample, every member has an equal chance of selection, so the sample is likely to resemble the population. A stratified sample divides the population into groups (strata) and samples from each in proportion. Convenience samples, such as asking people who walk past, are cheap but can be biased.
Worked example
A researcher wants to estimate the share of 10,000 students who work part-time. She selects 400 students at random and finds 160 working.
- Sample proportion = 160 ÷ 400 = 40%.
- A rough 95% margin of error for a random sample is about 1 ÷ √n = 1 ÷ 20 = 5 percentage points.
- So the true share is likely between about 35% and 45%.
A larger random sample narrows the margin, but quadrupling the sample only halves it. A sample of 1,600 would have a margin of about 2.5 points.
Ethics and cautions
Researchers must obtain informed consent, protect confidentiality and avoid harm. A correlation between two variables does not by itself prove that one causes the other. A spurious relationship appears when a third factor drives both.