Sampling: who can speak for whom?

By UK GCSE Sociology · independent study · myAudience

Original OpenAI imagegen conceptual teaching illustration, generated 4 October 2026; not observational evidence. · AI-generated original asset for this independent guide; no third-party artwork intentionally reproduced.

Sampling: who can speak for whom?

The target population is the group about which a study seeks to draw conclusions. A sampling frame is an operational list or source from which cases can be selected. A sample is the selected subset, while the achieved sample consists of those who actually contribute usable evidence. These distinctions expose different exclusions. If a study of local adult library users samples a list of newsletter subscribers, users who never subscribe are absent before selection begins. If selected subscribers do not respond, further differences can appear. A large achieved sample can still systematically omit important parts of the target population.

In simple random sampling, every member of a complete frame has an equal known chance of selection through a genuinely random process. Random does not mean whoever happens to be nearby, nor does it guarantee that one realised sample exactly reproduces every population characteristic. Systematic sampling selects at a fixed interval after an appropriate starting choice; periodic patterns in the list can create bias. Stratified sampling divides the frame into relevant groups before selection, often with random selection within each group. The definition of strata and their proportions matter. Stratifying cannot include people missing from the frame or prevent unequal response rates.

Quota sampling seeks specified numbers in categories but usually leaves selection within categories to the recruiter. Matching a few proportions does not make the sample random or representative in every respect. Convenience or opportunity sampling recruits accessible cases. Snowball sampling begins with contacts who suggest further participants; it can help reach networks without a usable frame, but disconnected people and less connected networks may be missed. Purposive selection can deliberately seek cases relevant to a qualitative question. Judge the approach against the aim: exploring a range of experiences is different from estimating the proportion of a whole population holding a view.

Non-response concerns selected cases who do not supply data. If those with inconvenient opening hours are less likely to answer a library survey, the achieved sample may overstate satisfaction. Record the number invited and the number responding, and examine whether non-response could relate to the topic. Increasing the number invited may improve precision but does not automatically remove bias. Weighting can sometimes adjust known imbalances, but it cannot recover unobserved experiences without assumptions. At GCSE level, the crucial reasoning is to locate the selection problem and explain its likely effect on a particular conclusion.

Representativeness concerns how well the sample reflects relevant features of the population for the inference being made. Reliability concerns consistency of a procedure or measure. Validity concerns whether evidence captures what it is intended to capture. A repeatable survey of newsletter subscribers may be reliable yet weak for generalising to all users. A careful interview may validly illuminate one person's experience without estimating its prevalence. Keep these judgments separate. Generalisation is a reasoned step from evidence to a wider claim, not a reward automatically earned by a high participant count or an official institution.

Read the token illustration as a metaphor: the broad tray suggests a population, the accessible tray a frame, the bowl a sample and the partly hidden tray possible exclusions. Colors and shapes are not demographic categories, and their proportions are not calibrated. The picture cannot demonstrate probability selection. Every sampling example in this lesson is fictional. Ethical access also constrains sampling: inclusion in a frame is not consent, and a gatekeeper's permission does not remove a participant's right to decline. Lesson 08 examines that difference before any real project could be considered.

Worked example: Original hypothetical data: a library has 600 registered adult users, of whom 240 subscribe to its newsletter. A researcher randomly selects 120 subscribers; 72 reply and 54 say the hours suit them. Evaluate the claim that 75% of all registered adult users are satisfied.

Identify population 600, accessible subscriber frame 240, selected sample 120 and achieved sample 72. These are different denominators with different meanings.

Calculate satisfaction among respondents: 54/72 × 100 = 75%. Calculate response rate among invited subscribers: 72/120 × 100 = 60%.

Locate frame coverage: 240/600 = 40% of registered adult users subscribe. Random selection within that frame does not represent the 360 non-subscribers by itself.

Consider topic-related non-response and compare possible experiences of excluded users. Recommend a broader frame and accessible invitations, while keeping participation voluntary.

Answer: The arithmetic supports 75% satisfaction among the 72 respondents, not among all 600 registered adult users. The frame excludes non-subscribers, and 48 selected subscribers did not reply. Satisfaction or difficulty using the hours could influence both subscription and response. Random selection within the subscriber list reduces recruiter choice at that stage but cannot fix frame exclusion or non-response. A broader sampling frame and follow-up invitations could improve coverage, yet the final report would still need to discuss response patterns and how satisfaction was measured.

Check your understanding: Transfer task: a fictional college wants to estimate all enrolled learners' views of its canteen. Compare an online voluntary poll, quotas filled at lunchtime and a stratified random selection from enrolment records. Explain a remaining weakness of the strongest design. Indicative rubric: 12 points: each comparison 3, justified choice and remaining weakness 3.

Answer: The online poll has self-selection and possible digital-access differences. Lunchtime quotas can match chosen categories but omit learners absent at that time and retain recruiter discretion. Stratified random selection can cover enrolment groups if the frame is current, making it more defensible for population estimation. However, differential non-response or an unclear question can still distort results. The answer should also note that enrolled learners who never use the canteen need a meaningful response option.

Why: Award full comparison credit for identifying the selection process and its consequence for this target population. 'Random is best' without conditions is insufficient. The remaining weakness must survive the proposed improvement. A qualitative aim about why particular learners avoid the canteen could justify a different selection strategy, but that would answer a different question from estimating all learners' views.