⚫ Scatter Graphs and Correlation

GCSE Maths · Statistics (S6)

Ages 15–16 · Foundation & Higher

← Back to topic overview
1 What a Scatter Graph Shows
A scatter graph plots two variables against each other, one point per item. It shows whether the two variables are related — this is called bivariate data.
Never join the points on a scatter graph. Each point is a separate item, not a stage in a sequence. If a line is wanted, it is a single straight line of best fit.
2 Types of Correlation
Positive as x rises, y rises Negative as x rises, y falls No correlation no clear pattern Strong positive points close to a line
CorrelationMeaningExample
PositiveAs one increases, the other increasesHeight and shoe size
NegativeAs one increases, the other decreasesAge of a car and its value
NoneNo clear patternShoe size and exam score
StrongPoints lie close to a straight line
WeakPoints are scattered but show a tendency
Describing correlation
Always give both words: strength and direction
e.g. "strong negative correlation", not just "negative"
Worked Example 1 — Describing correlation

A scatter graph of a car's age against its value shows points falling steeply and lying close to a straight line. Describe the correlation and what it means.

The points fall from left to right, so the correlation is negative.
They lie close to a line, so it is strong.

Strong negative correlation: as a car gets older, its value tends to decrease.

The interpretation sentence must use the context — "as age increases, value decreases" — not just the word "negative".
3 The Line of Best Fit
outlier — ignore it read up… …then across Hours of revision Test score
Worked Example 2 — Using the line to predict

A line of best fit passes through $(2,\ 30)$ and $(8,\ 66)$ on a graph of revision hours against test score. Estimate the score for $5$ hours of revision.

Gradient $= \dfrac{66 - 30}{8 - 2} = \dfrac{36}{6} = 6$
Equation: $y = 6x + c$. Using $(2, 30)$: $30 = 12 + c$, so $c = 18$.
$y = 6x + 18$
At $x = 5$: $y = 30 + 18 = 48$

Estimated score: $48$ marks.

Worked Example 3 — Interpreting the gradient and intercept

Using $y = 6x + 18$ from above, interpret the gradient and the intercept in context.

Gradient $6$: for each extra hour of revision, the test score increases by about $6$ marks on average.
Intercept $18$: a student who does no revision would be predicted to score about $18$ marks.
Interpreting the gradient in context — "per extra hour" — is a very common exam question and needs the units.
4 Interpolation and Extrapolation
The two words
Interpolation — estimating inside the range of the data. Reliable.
Extrapolation — estimating outside the range. Unreliable.
Extrapolation is dangerous. The pattern seen in the data may not continue beyond it. The relationship might level off, reverse, or stop applying altogether — and there is no evidence either way, because no data were collected there.
Worked Example 4 — Judging a prediction

A scatter graph shows revision hours from $1$ to $9$ against test scores. Comment on the reliability of estimates for (a) $6$ hours, (b) $20$ hours.

(a) $6$ hours is within the range $1$–$9$, so this is interpolation — the estimate is reliable.
(b) $20$ hours is far outside the range, so this is extrapolation — unreliable.
In fact the line would predict a score above $100\%$, which is impossible. Beyond a point, extra revision brings diminishing returns.
Worked Example 5 — When extrapolation gives nonsense

A line of best fit for children's heights against age is $h = 6a + 75$, fitted using ages $4$ to $11$. Use it to predict the height of a $40$-year-old, and comment.

$h = 6(40) + 75 = 240 + 75 = 315$ cm
That is over three metres tall — clearly impossible.

The prediction is meaningless because $40$ is far outside the age range $4$–$11$ used to fit the line. People stop growing in their late teens, so the linear relationship simply does not hold at that age.

5 Correlation Is Not Causation
The most important sentence on this page
Correlation shows that two things vary together.
It does not show that one causes the other.
When two variables correlate, there are four possible explanations:
• $A$ causes $B$
• $B$ causes $A$
• A third factor causes both (a confounding variable)
• It is pure coincidence
Correlation observedThe real explanation
Ice cream sales and drownings both riseHot weather causes both
Shoe size and reading ability in childrenAge causes both
Number of firefighters and damage at a fireThe size of the fire causes both
Chocolate consumption and Nobel prizes by countryNational wealth is linked to both
Worked Example 6 — Explaining a spurious correlation

A study finds a strong positive correlation between the number of ice creams sold and the number of people who get sunburnt. Does eating ice cream cause sunburn?

No. There is a correlation, but no causal link between the two.
A third factor — hot, sunny weather — causes both. On hot sunny days more people buy ice cream and more people are outside getting sunburnt.

This third factor is called a confounding variable.

Worked Example 7 — When causation might be reasonable

A scatter graph shows strong negative correlation between the amount of insulation in a house and its annual heating bill. Is it reasonable to suggest causation here?

There is a plausible mechanism: insulation reduces heat loss, so less energy is needed.
The direction makes sense: insulation is installed first, and the bill follows.

Causation is plausible here — but the scatter graph alone still does not prove it. Other factors (house size, how warm people keep their homes) could also be involved. A controlled experiment would be needed to be sure.

Correlation plus a credible mechanism is stronger evidence than correlation alone — but it is still not proof.
6 Outliers on Scatter Graphs
An outlier on a scatter graph is a point that lies well away from the general pattern. It may be a recording error, or a genuine unusual case.
Worked Example 8 — Dealing with an outlier

On a graph of revision hours against test score, most points follow a rising pattern, but one student who revised for $2$ hours scored $95$. How should this be treated?

Identify it as an outlier — it does not follow the pattern of the other points.
Ignore it when drawing the line of best fit, so the line represents the general trend.
Possible explanations: the student may already have known the material, or the data may have been recorded wrongly.

Do not simply delete the point from the graph — plot it, but do not let it pull the line.

7 Quick Reference

Scatter graph

One point per item; never join the points.

Positive

Both variables increase together.

Negative

One increases as the other decreases.

Describe it fully

Strength and direction, then interpret in context.

Line of best fit

Straight, follows the trend, roughly equal points either side.

Gradient

The change in $y$ per unit increase in $x$ — give the units.

Interpolation

Inside the data range — reliable.

Extrapolation

Outside the data range — unreliable.

Correlation

Never proves causation; look for a third factor.

Outliers

Plot them, but ignore them when drawing the line.

8 Practice Questions
Question 1

Describe the correlation you would expect between: (a) a car's engine size and its fuel economy, (b) a person's height and their arm span, (c) a person's house number and their age.

▶ Show solution

(a) Negative — larger engines generally use more fuel, so economy falls.

(b) Strong positive — taller people generally have longer arm spans.

(c) No correlation — the two are unrelated.

Question 2

Explain why you should never join the points on a scatter graph.

▶ Show solution

Each point represents a separate item (a different person, car or day), not consecutive stages of one thing changing.

Joining them would suggest a sequence or a path between the points, which has no meaning. If a line is needed, a single straight line of best fit is drawn instead.

Question 3

A line of best fit passes through $(0,\ 12)$ and $(10,\ 52)$. Find its equation.

▶ Show solution

Gradient $= \dfrac{52 - 12}{10 - 0} = \dfrac{40}{10} = 4$

The line crosses the vertical axis at $12$, so $c = 12$.

$y = 4x + 12$

Question 4

Using $y = 4x + 12$ from Question 3, where $x$ is hours worked and $y$ is pay in pounds, interpret the gradient and the intercept.

▶ Show solution

Gradient $4$: pay increases by £$4$ for every extra hour worked.

Intercept $12$: a fixed payment of £$12$ is received even for zero hours — perhaps a call-out fee or bonus.

Question 5

Data were collected for ages $5$ to $16$. Explain whether an estimate for age $12$ and one for age $30$ would be reliable.

▶ Show solution

Age $12$: inside the range $5$–$16$, so this is interpolation — the estimate is reliable.

Age $30$: well outside the range, so this is extrapolation — unreliable, because there is no evidence that the pattern continues beyond age $16$.

Question 6

A study finds a positive correlation between the number of storks nesting in a region and the number of babies born there. Does this mean storks deliver babies? Explain.

▶ Show solution

No. The correlation is real but there is no causal link.

A likely third factor is the size or type of the region: larger rural areas have both more buildings for storks to nest on and more people, and therefore more births.

This is a classic example of correlation without causation.

Question 7

A line of best fit on a graph of a car's age (years) against value (£) is $V = 15\,000 - 1200a$. Interpret the gradient and use the equation to estimate the value of a $6$-year-old car.

▶ Show solution

Gradient $-1200$: the car loses about £$1200$ of value each year.

At $a = 6$: $V = 15\,000 - 1200(6) = 15\,000 - 7200 = £7800$

Question 8

Using the equation from Question 7, find the value predicted for a $15$-year-old car, and comment.

▶ Show solution

$V = 15\,000 - 1200(15) = 15\,000 - 18\,000 = -£3000$

A negative value is impossible, so the prediction is meaningless.

This shows the danger of extrapolating too far: in reality a car's value falls steeply at first and then levels off towards a small positive amount, rather than continuing to drop by £$1200$ every year.

Question 9

A scatter graph of temperature (°C) against ice cream sales shows a line of best fit through $(10,\ 40)$ and $(25,\ 130)$.

(a) Find the gradient and interpret it.   (b) Find the equation.   (c) Estimate sales at $18$°C.   (d) Would you use the line to estimate sales at $35$°C?

▶ Show solution

(a) Gradient $= \dfrac{130 - 40}{25 - 10} = \dfrac{90}{15} = 6$

Sales increase by about $6$ ice creams for each $1$°C rise in temperature.

(b) $y = 6x + c$; using $(10, 40)$: $40 = 60 + c$, so $c = -20$.

$y = 6x - 20$

(c) At $x = 18$: $y = 108 - 20 = 88$ ice creams.

(d) $35$°C is outside the data range $10$–$25$°C, so this would be extrapolation and unreliable. The estimate of $190$ might also exceed what the shop can physically sell in a day.

Question 10

The table shows the hours of sleep and reaction times (milliseconds) for eight people.

Sleep (h)$4$$5$$5.5$$6$$7$$7.5$$8$$9$
Reaction (ms)$420$$390$$385$$360$$330$$325$$300$$285$

(a) Describe the correlation.   (b) A line of best fit passes through $(4,\ 420)$ and $(9,\ 285)$. Find its equation.   (c) Estimate the reaction time for someone who sleeps $6.5$ hours.   (d) A newspaper claims "sleeping more makes you react faster". Comment on this claim.

▶ Show solution

(a) As sleep increases, reaction time decreases, and the points follow the pattern closely.

This is strong negative correlation — people who sleep longer tend to have faster reaction times.

(b) Gradient $= \dfrac{285 - 420}{9 - 4} = \dfrac{-135}{5} = -27$

Using $(4, 420)$: $420 = -27(4) + c = -108 + c$, so $c = 528$.

$y = -27x + 528$

(c) At $x = 6.5$: $y = -27(6.5) + 528 = -175.5 + 528 = 352.5$ ms

(d) The data show a strong correlation, but that does not establish causation.

It is plausible that more sleep improves alertness, so the claim is not unreasonable — but a third factor could explain the pattern. For example, people who are ill, stressed or older may both sleep less and react more slowly. To support the claim properly you would need a controlled experiment in which sleep is varied deliberately.

Scatter Graphs & Correlation (S6) · GCSE Maths Revision · Created with MathJax