A statistical investigation is a connected cycle: question → population → sample → collect → analyse → infer → report. A weakness early in the cycle can limit every conclusion that follows.
The question should identify the population and variable clearly. Sampling may be necessary because a census is too expensive, slow or restricted by access. A fair sample should give relevant population members a reasonable chance of representation.
Ethical practice includes consent where appropriate, privacy/de-identification, respectful wording, safe data storage and avoiding unnecessary collection of sensitive information. Fairness is both a statistical and ethical issue.
Analyse the sample using graphs and summaries that suit the data. Then make a population inference in cautious language: “the sample suggests…” rather than “this proves…”. Acknowledge sampling variation, possible bias, measurement limits and context.
Preserve the authored applications: electricity-consumption data around major events such as pandemics; choosing samples because of efficiency, cost or restricted time; and investigating progress in reconciliation between First Nations Australians and non-Indigenous Australians using carefully evaluated sampling and data sources.