1.2 - Populations and sampling
Researchers usually study a sample rather than everyone they want to understand. Learn how four sampling techniques recruit participants, why random selection differs from convenient recruitment, and how sampling affects the scope of a conclusion.
Target population and sample
The target population is the group to which researchers want their findings to apply.
A study could aim to understand all first-year students at a particular college. Those students are its target population. The sample is the subset actually studied, such as 30 students who complete a questionnaire. Specify the intended population before recruitment; “everyone” is rarely a defensible target for a small local study.
A representative sample reflects relevant characteristics of that population sufficiently well for the purpose of the study. If evening students have different pressures from daytime students, studying only daytime students may distort conclusions about the whole college. Representativeness concerns relevant characteristics, not just whether the sample is large. Including participants of different ages does not automatically make a sample representative of a target defined by several other features.
Generalisability concerns applying findings beyond the studied participants. Sampling evidence supports a particular scope; one college does not automatically represent all students nationally. The method, task and setting matter too, so sampling alone does not guarantee generalisability. Participation must be voluntary under every technique; a selection procedure does not give permission to compel someone to take part.
Random sampling
In a simple random sample, every member of the defined target population has an equal chance of selection.
Researchers need an appropriate sampling frame, such as a complete current list of the college's 200 first-year students. Number each student once, then use a random-number procedure to select 20 distinct numbers without replacement. Invitations go to those selected. Drawing names from a thoroughly mixed container is another random procedure if each eligible person appears once.
This reduces deliberate researcher selection bias and gives every listed person an opportunity to be selected. It does not guarantee a representative mix in one small sample. Chance variation can overrepresent a group, and people who decline may differ from those who accept. A frame that omits evening students gives them zero selection chance, so random numbers applied to that incomplete list do not create a random sample of the full target population.
For a fixed sample of 20 from 200, each listed person's selection probability is , or 10%. Equal selection chances are the important principle; the procedure does not promise equal numbers from every subgroup. Obtaining a complete frame, contacting selected people and handling non-response can take time.
Random sampling selects people from the population. Random allocation assigns recruited participants to experimental conditions. A study can use an opportunity sample and then randomly allocate its members; that does not make its recruitment random.
Opportunity and self-selected samples
Opportunity and self-selected sampling both offer convenient recruitment but differ in how participation begins.
In opportunity sampling, the researcher recruits available and willing people, for example approaching students passing a campus café at lunchtime. This is quick and inexpensive and does not need a full population list. However, who happens to be available depends on the place and time. Evening students or those bringing lunch from home may be absent, so café recruits may not represent all first-year students.
In self-selected sampling, also called volunteer sampling, people choose to respond to a recruitment invitation. A college-wide notice could invite students to contact researchers about a wellbeing survey. This can reach more locations than approaching one café, and respondents are willing to participate. Yet the invitation reaches only people who see it, and those who volunteer may be unusually interested in wellbeing or have strong experiences to report. Willingness does not guarantee typicality.
The distinction is the recruitment procedure, not merely consent: opportunity participants also agree voluntarily. If a researcher approaches available passers-by, it is opportunity sampling; if participants initiate contact after an advertisement, it is self-selected. Different routes can occur in one study, in which case report them accurately rather than force a single label.
Using notices in several places or approaching people at different times can broaden access. These changes reduce a particular exclusion but do not give each population member the known equal chance of a simple random sample.
Snowball sampling
Snowball sampling recruits initial participants and asks them to help reach further eligible people through their contacts.
A researcher investigating the experiences of recently arrived students who use an uncommon first language might start with a few eligible volunteers, then ask them to pass an invitation to other eligible people. Those people can pass it on again. This is useful when the target group is hard to locate and a complete list is unavailable. A trusted invitation through a contact can also make an unfamiliar research project more approachable.
The chain follows social networks, so closely connected people may be overrepresented while isolated people are missed. People often know others with similar backgrounds or experiences; expanding the number recruited from one network does not necessarily broaden the perspectives. Starting several separate chains may improve reach, but recruitment remains non-random.
Researchers should protect privacy. An initial participant should not be pressured to reveal another person's sensitive status or contact details without permission. Passing on an invitation lets a new person decide whether to approach the researcher. The researcher still checks eligibility and seeks that person's consent; a friend's recommendation does not substitute for consent.
Judging sampling evidence
A sampling judgement links who was recruited to the particular conclusion a researcher wants to draw.
Use the population total when describing population composition and the sample total when describing sample composition.
Suppose the target population has 120 daytime and 80 evening students, a total of 200. Daytime students form of that population. A café opportunity sample has 18 daytime and 2 evening students. Daytime students form of this sample. It overrepresents daytime students on this relevant characteristic; the two percentages describe composition, not proof that every opinion is biased.
For a random sample of 20, the expected daytime count is , but an actual random sample need not contain exactly 12. Chance composition differs from a recruitment procedure that systematically excludes evening students. A larger random sample generally reduces random sampling fluctuation; a larger lunchtime-only opportunity sample can preserve a systematic exclusion.
To improve the café study, recruit at evening as well as daytime sessions and check which groups still fail to respond. If a complete frame is available, random selection offers a clearer selection process. Report the frame, recruitment procedure, numbers invited and participating, and relevant sample characteristics so the reader can judge its limits.
The best technique depends on access and purpose. Random sampling needs a usable frame, opportunity sampling is practical for a small initial study, self-selection can reach willing volunteers, and snowball sampling can locate a hard-to-reach group. None makes a sample representative by its name alone. Avoid claiming national conclusions from a narrow local sample without further evidence.