Social research methods, measurement and evidence
Research methods help assess whether evidence supports a social or policy claim. Begin with a question, the population or situation of interest and the observation that would bear on it. Distinguish an abstract concept from the indicator used to measure it. Public participation, for example, is not automatically represented by one count of visits if the intended concept also includes who participated, what they did and which groups were excluded.
Sampling and inference
A sample is not representative merely because it has a large number of responses. Examine how participants entered the study, who was eligible and who could not be reached. Sampling-frame omissions differ from nonresponse among people who were invited. A convenient group in one affluent neighbourhood cannot establish a universal claim about all residents. A stratified random design samples within each defined stratum. In a one-stage cluster design, clusters are selected and all members of those selected clusters are surveyed. Multistage designs may sample further within selected clusters; complete enumeration is not a universal definition of every cluster design.
Measures and questions
Reliability concerns consistency of a measurement procedure, while validity concerns the interpretation the evidence supports. Agreement between coders is different from agreement that a policy is desirable. A questionnaire that asks whether streets are clean and neighbours are friendly combines two judgments in one response. Split such a question so that a person can report different answers. Avoid wording that embeds the desired conclusion as an established premise. A pilot can reveal confusing wording or workflow problems without guaranteeing that every later source of error has disappeared.
Explanation and alternatives
A correlation does not by itself establish a causal effect. Compare the proposed explanation with plausible alternatives, including selection, a changing definition and a third variable. An aggregate association between regions does not determine the relationship for every individual within them. A qualitative interview can illuminate a participant's account without proving its frequency in the whole population. Combining sources is useful when their different assumptions and limitations are examined, not when agreement is treated as automatic proof.
Present a defensible conclusion
State the relevant evidence, the reasoning linking it to the claim and the limit of that inference. Distinguish a policy objective from an observed outcome and an absence of demonstrated harm from proof of universal safety. Identify what further data could change the conclusion. Protect participants and interpret the supplied consent and privacy conditions without inventing a national statutory procedure. These methods support supplied-material policy analysis and argument evaluation; they do not substitute a biology curriculum, current political doctrine or an official timed writing assessment.