Fieldwork Sampling Judgement
A student conducting fieldwork for a tourism investigation surveys visitors at a popular lookout. She stands at the exit between 10am and 12pm on a Saturday and approaches every fifth person, achieving 48 responses. She plans to generalise: 'Most visitors to the region are day-trippers who stay less than four hours.' Her data shows 62% of respondents stayed under four hours. A classmate points out that the survey missed overnight guests who leave early, and that the lookout is mainly a day-trip stop. The student argues her sample is 'random because I used every fifth person'.
1. Evaluate the student's claim that her sample is random, and identify the sampling problems.
Read the worked answer
The sample is not random in the statistical sense. 'Every fifth person' is systematic sampling, which can be acceptable, but randomness requires that every member of the population has an equal chance of selection. Here, the population is 'visitors to the region', but the sampling frame is 'people exiting this lookout between 10am and 12pm on a Saturday'.
Problems:
- Time bias: a two-hour Saturday window excludes overnight guests, weekday visitors, and early or late visitors.
- Location bias: a day-trip lookout over-represents day-trippers by design.
- Coverage bias: the sample cannot represent visitors who never come to this lookout.
- Systematic selection may also interact with group sizes (e.g., every fifth person in a couple or family).
So the method is systematic but not representative, which is the key distinction.
2. Assess whether the data supports the generalisation 'most visitors to the region are day-trippers who stay less than four hours'.
Read the worked answer
The generalisation is not supported for the region as a whole.
The 62% figure describes respondents at this lookout in this window, not all regional visitors. Because the location and timing structurally favour day-trippers, the sample is biased toward the very conclusion drawn - a form of selection effect. Overnight guests who leave early or visit elsewhere are absent, so their stay length is invisible in the data.
A defensible statement would be: 'Among visitors surveyed at this lookout on a Saturday morning, 62% stayed under four hours; this cannot be generalised to all regional visitors because the sampling excluded other locations and times.'
The exam rewards recognising the difference between what the data shows and what the student wants it to show.
3. Recommend two changes to the sampling design that would make the generalisation more defensible, and explain the trade-off.
Read the worked answer
Two changes:
- Sample across multiple sites and times - e.g., accommodation areas, transport hubs, and the lookout, on weekdays and weekends, morning and evening - so overnight and day visitors are both captured.
- Stratify the sample - deliberately sample subgroups (overnight vs day visitors, different accommodation types) in proportion to their estimated share, then weight results.
Trade-off: broader sampling costs time, access and resources, and may reduce response numbers at each site. Stratification requires reliable estimates of subgroup sizes, which may not exist. The student must therefore balance representativeness against feasibility and should acknowledge remaining limitations rather than claim full representativeness.
The judgement is that better design narrows the gap between sample and population but rarely eliminates it - so conclusions should stay proportionate to the evidence.