MKTG90011 · Marketing Research
Sampling
A sample is the subset you actually study, and how well it stands in for the population depends on the sample design. The first fork is everything: probability methods give every unit a known, non-zero chance of selection — so you can generalise to the population and compute sampling error — while non-probability methods (convenience, judgement, quota, snowball) are cheap and fast but cannot be projected to the population. The sampling process is fixed: define the target population, obtain a sampling frame (a list of every unit — note frame ≠ population, which causes frame error), choose a method, determine the sample size, and execute. Sample size is computed from the confidence level, the margin of error, and the population variability (σ or a proportion). Finally, keep two error types straight: sampling error (the luck of which units you drew, shrinks with n) versus non-sampling error (bad questions, non-response, data-entry mistakes — which a bigger sample does not fix).
What this chapter covers
- 016.1 Population, sample, sampling frame and parameter vs statistic
- 02The five-step sampling process
- 03Probability methods: simple random, systematic, stratified, cluster
- 04Non-probability methods: convenience, judgement, quota, snowball
- 05Generalisability — why only probability samples project
- 06Sample-size determination (confidence, margin of error, σ/proportion)
- 07Sampling vs non-sampling error
Worked example: name the method and judge generalisability
- +1(a) Every 10th name → systematic sampling (probability). A known selection interval gives every unit a known chance, so results can be generalised (watch for periodicity in the list).
- +1(b) Whoever walks past → convenience sampling (non-probability). Selection chance is unknown and self-selecting, so results cannot be projected to the population.
- +1(c) Referrals from respondents → snowball sampling (non-probability). Useful for hard-to-reach groups but non-generalisable.
- +1(d) Random within proportional age strata → stratified sampling (probability). Generalisable, and more precise than simple random when strata differ.
Key terms
- Population vs sample
- The population is the entire group of interest, described by a parameter (e.g. μ); the sample is the subset actually examined, described by a statistic. Inference uses the statistic to estimate the parameter.
- Sampling frame
- The list of every unit in the population from which the sample is drawn. When the frame doesn't match the population (a phone directory missing mobile-only households) it causes frame error.
- Probability sampling
- Methods (simple random, systematic, stratified, cluster) in which every unit has a known, non-zero chance of selection — the only family that supports generalising to the population and computing sampling error.
- Non-probability sampling
- Methods (convenience, judgement, quota, snowball) where selection chance is unknown. Cheap and fast, useful for exploratory or hard-to-reach work, but results cannot be projected to the population.
- Sampling vs non-sampling error
- Sampling error is the random difference between sample and population from which units were drawn — it shrinks as n grows. Non-sampling error (bad questions, non-response, coding mistakes) is not fixed by a larger sample.
Sampling FAQ
What is the key difference between probability and non-probability sampling?
Probability sampling gives every unit a known, non-zero chance of selection, so you can generalise to the population and compute sampling error. Non-probability sampling (convenience, judgement, quota, snowball) has unknown selection chances — it is cheap and fast but its results cannot be projected to the population.
What determines the sample size I need?
Three inputs: the confidence level you want (e.g. 95%), the margin of error you can tolerate, and the variability in the population (σ for a mean, or a proportion). More confidence or a tighter margin or more variability all push the required n up; the formula combines them.
What is the difference between sampling error and non-sampling error?
Sampling error is the random gap between your sample and the population caused by which units you happened to draw — it shrinks as the sample grows. Non-sampling error comes from flawed questions, non-response, interviewer or data-entry mistakes, and a bigger sample does not fix it — better design does.
Why does the sampling frame matter?
Because the frame — the actual list you sample from — is rarely identical to the population. If the frame omits part of the population (frame error) your sample is biased before you start, no matter how good your method. Always check the frame matches the target population.
Exam move
Build the one grid that answers most sampling items: method → probability or not → generalisable or not, and learn to read the method from its mechanism (interval = systematic, referrals = snowball, random-within-strata = stratified). Be ready to compute n from confidence, margin of error and variability with the permitted calculator, and to distinguish sampling vs non-sampling error (only the former shrinks with n) — both are recurring marks. Keep frame ≠ population in mind, because frame error is a favourite trap.
