Exploring and designing practice questions.

A student proposes the research question 'Does farming harm the river?'. For an ESS investigation the most useful improvement is to

• name the variable measured, the comparison made and the sites used
• add a prediction stating that nitrate concentration will be higher downstream of the fields
• widen the question so that more types of farming can be included
• replace 'harm' with a less emotive word such as 'affect'

Answer: name the variable measured, the comparison made and the sites used

An operationalised research question states exactly what will be measured, in what units, and between which places or treatments the comparison is made. A prediction is useful but belongs in the hypothesis, and it cannot rescue a question that has no measurable variable.

A hillside grassland contains three clearly different soil types covering unequal areas, and a student needs one mean earthworm density for the whole hillside. The most appropriate sampling strategy is

• systematic sampling at fixed 10 m intervals along a transect down the slope
• opportunistic sampling wherever earthworm casts can be seen
• stratified sampling with quadrat numbers proportional to each soil area
• random sampling within the largest soil type only, then scaling up

Answer: stratified sampling with quadrat numbers proportional to each soil area

Stratified sampling with effort shared out in proportion to area guarantees that each soil type contributes to the mean according to how much of the hillside it covers. A transect would sample the slope gradient rather than the soil pattern, and sampling only the largest soil type builds bias into the estimate.

A student investigates how the mass of nitrate fertiliser added affects the growth of duckweed in identical tanks. Which of these is a controlled variable?

• The mass of nitrate added to each tank
• The light intensity above every tank
• The increase in duckweed surface cover
• The difference in growth between the tanks

Answer: The light intensity above every tank

Controlled variables are the ones deliberately held constant so that only the independent variable can account for a difference in the outcome. Here nitrate mass is the independent variable and duckweed cover is the dependent variable, so neither can also be controlled.

Before the main fieldwork a student spends one afternoon carrying out a pilot study. The main purpose of a pilot study is to

• supply the raw data that will be processed as the final result
• increase the sample size so that the mean becomes more reliable
• prove the hypothesis before the full investigation begins
• check the method and refine the range of values chosen

Answer: check the method and refine the range of values chosen

A pilot tests whether the method is workable and whether the chosen range of the independent variable produces a measurable response, so that the design can be changed before time is wasted. Its data are diagnostic and are not reported as results.

A student compares invertebrate diversity in a stream above and below a sewage outfall. Which design feature most directly reduces the effect of random variation on the two means?

• Taking five replicate kick samples at each of the two sites
• Using the same net and the same three minute sampling time at both sites
• Recording the results in a table with the units in the headings
• Choosing two sites that are 200 m apart on the same stream

Answer: Taking five replicate kick samples at each of the two sites

Replication lets unpredictable variation between samples average out, so the mean sits closer to the true value and its spread can be quantified. Standardising the net and the timing removes systematic differences between sites, which is a different problem.

A student draws five water samples from the same spot in a stream within a minute of one another and treats them as five replicates for that site. The problem with the design is that

• the five samples describe one point rather than five independent points
• five replicates are always too few for a valid comparison
• water samples cannot be replicated at all in field studies
• the samples should have been taken from five entirely different rivers instead

Answer: the five samples describe one point rather than five independent points

Replicates have to be independent, meaning they could have differed for the reasons the study is about. Five bottles filled at one spot measure the precision of the sampling, not the variability of the site, and treating them as replicates makes the estimate look far more certain than it is.