Understand the idea.
A pattern is not a mechanism
An association means two measurements vary together. A causal claim says changing one produces a change in the other. Observational data show what happened together, but do not by themselves isolate what would happen if only one factor were changed.
A confounder may influence both variables. Reverse influence, selection differences and chance can also produce patterns. Offer plausible alternatives as possibilities to investigate, rather than presenting an unmeasured third variable as an established explanation.
Match the conclusion to the design
Random assignment can make treatment groups more comparable and strengthen causal evidence, especially with consistent measurement and comparable conditions. It differs from merely taking a random sample: sampling concerns representation; assignment concerns which treatment participants receive.
When only observational evidence is available, describe the association and its limits clearly. Consider other explanations and the study design. A strong pattern can motivate further study while still being insufficient to prove the claimed cause.
See the method in action.
Ice creams and sunburn
A beach survey finds that ice-cream purchases and sunburn cases both rise on some days. Does buying ice cream cause sunburn?
Worked solution
- 1
Identify what was observed.
Purchases and sunburn cases rise together on some surveyed days, which establishes an observed association.
- 2
Consider a common influence.
Sunny weather may increase ice-cream demand and sun exposure, plausibly affecting both measurements.
- 3
Limit the conclusion to the design.
The survey did not isolate the effect of buying ice cream, so it does not justify a causal conclusion.
No causal conclusion is justified; sunny weather is a plausible confounder.
Revision app comparison
Students choosing a revision app score higher than students not using it. Give an alternative explanation.
Worked solution
- 1
Identify how the groups formed.
Students chose whether to use the app, so the comparison involves self-selected groups.
- 2
Give a plausible alternative explanation.
App users may already be more motivated, have higher prior attainment, or study longer; these differences could contribute to their scores.
- 3
Describe stronger evidence.
Randomly assigning comparable students to app conditions, with the same assessment, would help separate the app's effect from pre-existing differences.
Self-selection and other study differences may explain the score association.
Try it yourself.
Check your understanding
Older children tend to have larger vocabularies and larger shoes. Does shoe growth cause vocabulary growth?
Show answer & reasoning
No.
Age and development plausibly affect both shoe size and vocabulary. Their association does not show that shoe growth causes vocabulary growth.
Keep practising
6 worksheet questions · 15 marks
Read the questions online
- Umbrella sales and rainfall rise together. Which is a plausible direction of causation?2 marks
- Towns with more fire engines report more total fire damage. Why is reducing fire engines not justified by this association?3 marks
- Coffee drinkers in a survey sleep less. Name a possible confounder.2 marks
- Why does random assignment help a treatment comparison?3 marks
- Write a cautious conclusion: more exercise is associated with lower reported stress in a voluntary survey.2 marks
- A company claims its product caused improvement because users improved over a month, with no comparison group. Give two alternative explanations.3 marks
Read the worked solutions
Equivalent correct methods earn the corresponding marks. Include units where asked.
- Umbrella sales and rainfall rise together. Which is a plausible direction of causation?2 marks
The context supports rain influencing purchases.
Sales alone do not cause weather.
Rainfall can increase demand for umbrellas. - Towns with more fire engines report more total fire damage. Why is reducing fire engines not justified by this association?3 marks
Fire severity can influence both engine numbers and damage.
The observed pattern does not isolate engines' protective effect.
Larger/more severe fires call out more engines. - Coffee drinkers in a survey sleep less. Name a possible confounder.2 marks
Shift work/workload may affect both coffee use and sleep.
The survey alone cannot separate these effects.
For example, shift work or workload. - Why does random assignment help a treatment comparison?3 marks
Participants do not choose their treatment.
Other influences are less likely to align systematically with treatment.
It reduces systematic group differences on average. - Write a cautious conclusion: more exercise is associated with lower reported stress in a voluntary survey.2 marks
Retain the observed direction.
State sample and design limits.
In this voluntary sample, more exercise is associated with lower reported stress; causation is not established. - A company claims its product caused improvement because users improved over a month, with no comparison group. Give two alternative explanations.3 marks
Time, other support or natural changes could contribute.
Without a suitable comparison, the product effect is not isolated.
Practice effects and other changes over the month, for example.
Curriculum reference & lesson scope
Australian Curriculum v 9 and SA Curriculum Mathematics
Year 10 continuous distributions, bivariate data, two-way categorical tables, bias and investigations, with the approved percentile and standard-deviation content included in the normal chapter.
Official curriculum source ↗