Quantitative and Qualitative Methods, and the Techniques of Social Research
2. Sampling
A sample is a subset drawn from a population (universe) according to a sampling frame, the list from which units are selected. In probability sampling every unit has a known, non-zero chance of selection, so sampling error can be estimated and results generalised. In non-probability sampling selection depends on judgement or convenience, which suits exploratory and qualitative work but does not permit statistical generalisation. Sampling error is the difference between a sample statistic and the population value arising from chance; it falls as sample size rises. Non-sampling errors — non-response, poor questions, interviewer bias — do not.
| Probability designs | Non-probability designs |
|---|---|
| Simple random — every unit equally likely, by lottery or random numbers | Convenience (accidental) — whoever is at hand |
| Systematic — every k-th unit after a random start, k = N/n | Purposive (judgemental) — units chosen for their relevance |
| Stratified — population divided into homogeneous strata (caste, gender), then random selection within each | Quota — strata filled to set numbers, but units chosen non-randomly |
| Cluster / multi-stage — naturally occurring groups (villages) sampled first, then units within them | Snowball — each respondent leads to others; used for hidden populations |
3. Questionnaire, schedule, observation, interview and case study
A questionnaire is a list of questions filled in by the respondent himself, usually sent by post or online; it is cheap and reaches a wide, dispersed, literate population, but suffers low response rates and cannot clarify misunderstood questions. A schedule is the same kind of list but filled in by the enumerator in a face-to-face meeting; it suits illiterate respondents, gives higher response and allows clarification, at greater cost and with the risk of interviewer influence. Questions may be open-ended or closed (fixed alternatives), and attitudes may be measured by scales such as the Likert scale (degrees of agreement), the Bogardus social distance scale and the Thurstone scale.
- Observation: participant (the researcher takes part in the group's life, openly or covertly) or non-participant; structured (with a checklist) or unstructured. Gold's four roles run from complete participant through participant-as-observer and observer-as-participant to complete observer.
- Interview: structured (fixed questions and order), semi-structured (an interview guide) or unstructured (conversational, in-depth). The focused interview of Merton, Fiske and Kendall (1956) concentrates on a respondent's experience of a known situation; the focus group interviews several people together to capture interaction.
- Case study: intensive study of a single unit — a person, family, village, organisation or event — in its full context, often using life histories and documents. Frédéric Le Play's family budget studies and Thomas and Znaniecki's The Polish Peasant in Europe and America (1918–20), built on letters and a life history, are classic examples.
4. Statistical analysis, interpretation and report writing
| Level of measurement (S.S. Stevens) | Property | Appropriate average and association |
|---|---|---|
| Nominal (religion, caste) | Categories only | Mode; chi-square test |
| Ordinal (class ranking, Likert response) | Categories with order | Median; Spearman's rank correlation |
| Interval (temperature in Celsius) | Equal intervals, arbitrary zero | Mean; Pearson's correlation |
| Ratio (income, age) | Equal intervals and a true zero | Mean, all ratios meaningful; Pearson's correlation |
Descriptive statistics summarise data — central tendency (mean, median, mode), dispersion (range, standard deviation) and association (correlation coefficients from −1 to +1, where correlation never by itself establishes causation). Inferential statistics test whether a sample result could have arisen by chance, as in the chi-square test for association between two categorical variables. Qualitative analysis proceeds by coding, grouping codes into themes, and comparing cases. Interpretation connects findings back to the research question and theory, and triangulation — using several methods, data sources or investigators on the same question — strengthens confidence. A research report conventionally sets out the problem, review of literature, objectives and hypotheses, methodology, findings, discussion and conclusions, with references and appendices containing the tools used.
Key takeaways
- Ethnography seeks the emic view through prolonged participant observation; surveys generalise from samples, cross-sectionally or longitudinally; the historical method weighs sources by external and internal criticism; the comparative method is, for Durkheim, sociology itself.
- Probability samples (simple random, systematic, stratified, cluster, multi-stage) permit estimation of sampling error; non-probability samples (convenience, purposive, quota, snowball) do not.
- A questionnaire is filled in by the respondent, a schedule by the enumerator; interviews range from structured to unstructured, and case studies examine one unit in depth.
- The level of measurement — nominal, ordinal, interval, ratio — decides which average and which test of association are appropriate.
- Correlation does not establish causation; triangulation strengthens interpretation; the report moves from problem and literature through method to findings and conclusions.
Practice questions (10)
Attempt each one before opening the answer. Every explanation names the tempting wrong option as well as the right one, because that is where marks are lost.
The basic difference between a questionnaire and a schedule is that:
Show answer
Answer: A — A questionnaire is filled in by the respondent, a schedule by the enumerator
Both are lists of questions and both can mix open and closed items; what separates them is who records the answers — the respondent (questionnaire) or the investigator in a face-to-face meeting (schedule).A researcher studying sex workers, a hard-to-reach population, asks each respondent to introduce others. This is:
Show answer
Answer: D — Snowball sampling
Snowball sampling builds the sample through referrals and suits hidden or stigmatised populations with no sampling frame; it is a non-probability design.Which of the following are probability sampling designs? Select all that apply.
Show answer
Answer: B — Multi-stage cluster sampling; C — Systematic sampling; D — Stratified random sampling
Systematic, multi-stage cluster and stratified random sampling all give every unit a known chance of selection. Quota sampling fills categories by the interviewer's choice and is non-probability.In systematic sampling from a list of 2,000 households, a sample of 100 is required. What is the sampling interval k?
Numerical answer — type the value.
Show answer
Answer: 20
The sampling interval is k = N/n = 2000/100 = 20: after a random start between 1 and 20, every twentieth household on the list is selected.Data on respondents' religion (Hindu, Muslim, Sikh, Christian and so on) are at which level of measurement?
Show answer
Answer: D — Nominal
Religious categories have no inherent order or distance between them, so the data are nominal; the mode is the appropriate average and chi-square an appropriate test of association.Establishing whether a document is genuine — its authorship, date and origin — is called:
Show answer
Answer: B — External criticism
External criticism tests authenticity; internal criticism then asks whether the content of an authentic document is accurate and credible. Both are steps of the historical method.The Polish Peasant in Europe and America, a classic of the life-history and documentary case-study approach, was written by:
Show answer
Answer: A — W.I. Thomas and Florian Znaniecki
Thomas and Znaniecki (1918–20) used letters, a long life history and other personal documents to study Polish migrants. Park and Burgess led the Chicago school's urban ecology; Goode and Hatt wrote Methods in Social Research; Glaser and Strauss developed grounded theory.Assertion (A): A strong correlation between two variables does not by itself show that one causes the other. Reason (R): Both variables may be influenced by a third variable. Choose the correct option.
Show answer
Answer: A — Both A and R are true, and R is the correct explanation of A
A spurious correlation produced by a common cause is precisely why correlation cannot establish causation; causal claims also need temporal order and the elimination of alternative explanations.Which pairing of method and landmark study is NOT correct?
Show answer
Answer: A — Panel study — Argonauts of the Western Pacific
Argonauts of the Western Pacific is Malinowski's ethnography of the Trobriand Islanders, not a panel study, which repeatedly surveys the same respondents over time.Using interviews, participant observation and official records together to study the same question, in order to cross-check findings, is known as:
Show answer
Answer: D — Triangulation
Triangulation, a term Norman Denzin systematised, combines methods, data sources, investigators or theories; convergence strengthens confidence and divergence prompts further inquiry.