Lesson 4 of 8 · 16 min
Describing associations in two-way tables
Use a percentaged two-way table to decide whether two categorical variables are associated, describe the association systematically with percentages in context, and evaluate claims.
What you will be able to do
- Decide whether there is an association by comparing percentages across the categories of the explanatory variable.
- Describe an association systematically and concisely, quoting the percentages that show it.
- Evaluate a claim made from a two-way table.
Two categorical variables are associated when the percentages for the response variable change from one category of the explanatory variable to another. Knowing someone's category then tells you something about how they are likely to respond.
If the percentages are about the same in every category, the explanatory variable tells you nothing new. There is no association.
Association
Age group and main news source (400 adults)
Online is the main source for 64% of 18–34-year-olds, 40% of 35–54-year-olds and 15% of those aged 55 and over.
The percentages change a lot from one age group to the next.
No association
Handedness and passing the driving test first time (700 learners)
61.3% of left-handed learners passed, compared with 59.7% of right-handed learners.
The percentages are almost the same, so handedness tells you nothing about passing.
Main news source within each age group. The online segment shrinks and the TV segment grows as age increases: an association.
Key idea
Is there an association?
Compare the percentages across the categories of the explanatory variable, one response category at a time.
- Clear differences (like 64%, 40% and 15%) suggest an association.
- Differences of only a few percentage points (like 61.3% and 59.7%) do not — differences that small can easily come from the luck of who happened to be surveyed.
There is no exact cut-off. Look at the size of the differences and say what you see.
A driving school recorded whether 700 learner drivers were left- or right-handed and whether they passed their practical driving test on the first attempt.
| Handedness | Passed | Did not pass | Total |
|---|---|---|---|
| Left-handed | 49 | 31 | 80 |
| Right-handed | 370 | 250 | 620 |
61.3% of left-handed learners passed, compared with 59.7% of right-handed learners. What does this suggest?
Describing an association systematically and concisely
Say whether there is an association, naming both variables
"There is an association between age group and main news source."
Quote percentages that show it
Choose the response categories where the differences are clearest, and give the percentage for every category of the explanatory variable.
Interpret the pattern in context
Say who is more likely to do what, in the words of the question.
Exam tip
A full-marks description
There is an association between age group and main news source. Younger adults are much more likely to get their news online: 64% of 18–34-year-olds, compared with 40% of 35–54-year-olds and 15% of those aged 55 and over. Older adults are more likely to rely on TV: 55% of those aged 55 and over, compared with 40% of 35–54-year-olds and 22% of 18–34-year-olds.
Worked example
Decide whether there is an association and describe it
A survey of 500 Queensland workers recorded their region and their main way of getting to work.
| Region | Car | Public transport | Walk or cycle | Total |
|---|---|---|---|---|
| Brisbane | 186 | 72 | 42 | 300 |
| Regional Queensland | 168 | 8 | 24 | 200 |
| Total | 354 | 80 | 66 | 500 |
Is there an association between region and main way of getting to work? Use appropriate percentages to support your answer.
Step 1: Choose the percentages
Region is the explanatory variable and it is in the rows, so calculate row percentages.
Step 2: Percentage the Brisbane row
car, public transport and walk or cycle.
Step 3: Percentage the regional row
car, public transport and walk or cycle.
Step 4: Compare across the regions
Region Car Public transport Walk or cycle Brisbane 62% 24% 14% Regional Queensland 84% 4% 12% Car use differs by 22 percentage points and public transport by 20. Walking or cycling is similar in both regions (14% and 12%).
Clear differences in any response category across the regions are evidence of an association. One category with similar percentages doesn't cancel that out.
Step 5: Write the conclusion in context
There is an association between region and main way of getting to work. Workers in regional Queensland are more likely to drive (84% compared with 62% of Brisbane workers), while Brisbane workers are much more likely to use public transport (24% compared with 4%). Similar percentages walk or cycle in both regions (14% and 12%).
Try it yourself first, then check the working one step at a time.
Answer
Yes. Regional workers are more likely to drive (84% compared with 62% of Brisbane workers) and Brisbane workers are more likely to use public transport (24% compared with 4%), so there is an association between region and main way of getting to work.
Method
Percentage within the categories of the explanatory variable.
Compare each response category across those categories, looking for clear differences.
State the association, quote the percentages for every category of the explanatory variable, and interpret in context.
Watch out for
Comparing counts: 186 Brisbane drivers against 168 regional drivers makes car use look similar, when the percentages (62% and 84%) are very different.
Common mistake
Descriptions that lose marks
- "There is a difference between Brisbane and regional workers." Too vague: which difference, and how big?
- "186 Brisbane workers drive compared with 168 regional workers." Counts, not percentages, so the different group sizes distort it.
- "Living in Brisbane makes people catch public transport." An association doesn't show that one variable causes the other.
Using the region and transport table from the worked example, which statement describes the association systematically and concisely?
QCAA questions often ask you to evaluate a claim. First check what the claim is based on. A claim built on counts from groups of different sizes can point the wrong way — calculate the percentages and see whether they support it.
Worked example
Evaluate a claim built on counts
A Townsville café owner recorded the milk chosen by 450 customers who ordered a coffee.
| Age group | Oat milk | Other milk or none | Total |
|---|---|---|---|
| Under 35 | 84 | 216 | 300 |
| 35 and over | 63 | 87 | 150 |
| Total | 147 | 303 | 450 |
The owner claims: "Younger customers are more likely to choose oat milk — 84 of them ordered it, compared with only 63 older customers." Evaluate the owner's claim.
Step 1: Identify what the claim is based on
The claim compares counts (84 and 63). The age groups are different sizes — 300 customers under 35 and only 150 aged 35 and over — so the counts can't be compared fairly.
A group twice the size will tend to have bigger counts in every column, whatever its preferences.
Step 2: Calculate percentages within each age group
Under 35: chose oat milk.
35 and over: chose oat milk.
Step 3: Compare
Customers aged 35 and over were more likely to choose oat milk (42% compared with 28%). The data suggests an association between age group and choosing oat milk — in the opposite direction to the claim.
Step 4: Evaluate the claim
The claim is not supported. It rests on counts, and younger customers placed more oat milk orders only because there were twice as many younger customers. The percentages show that older customers are more likely to choose oat milk.
The data also comes from one café's customers, so the pattern may not hold for other cafés.
"Evaluate" asks for a judgement with reasons: say whether the claim is reasonable, show the evidence, and note any limitation of the data.
Try it yourself first, then check the working one step at a time.
Answer
The claim is not supported: 42% of customers aged 35 and over chose oat milk, compared with only 28% of those under 35. Younger customers placed more oat milk orders only because there were twice as many of them.
Method
Check whether the claim uses counts or percentages, and whether the groups are the same size.
Calculate percentages within each category of the explanatory variable.
Decide whether the percentages support the claim, and say why.
Mention any limitation of the data, such as where the sample came from.
Watch out for
Agreeing with the claim because 84 is bigger than 63.
Percentaging down the oat milk column (), which only shows that most oat milk orders came from the bigger group.
Last week, 90 of a school's 150 Year 3 students borrowed a library book, as did 70 of its 100 Year 5 students. A teacher says Year 3 students are more likely to borrow books because 90 is more than 70.
What percentage of the Year 5 students borrowed a book?
Note
Association is not causation
An association tells you that two variables are linked, not why. Brisbane workers may use public transport more because there are more services, not because of where they live as such. You will look at non-causal explanations in Bivariate data analysis 2.
Summary
Two categorical variables are associated when the percentages for the response variable differ across the categories of the explanatory variable.
Similar percentages in every category mean there is no association.
Describe an association by naming both variables, quoting the percentages for every category of the explanatory variable, and interpreting the pattern in context.
To evaluate a claim, check whether it rests on counts or percentages, and whether the percentages support it.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.