Step-by-step worked examples and graded practice questions on correlation — bivariate data and scatter diagrams, describing correlation in context, why correlation does not imply causation, distinct sections of a population and outliers, and interpreting and using the regression line of \(y\) on \(x\) to make predictions. Written to the Edexcel Statistics Year 1 specification and equally suitable for AQA and OCR A.
📚 Statistics Year 1 (AS)✅ 15 Practice Questions🔍 7 Worked Examples⚠️ Common Mistakes
Struggling with scatter diagrams?
Alamin diagnoses exactly which skills are missing and builds a structured plan to fix them — backed by AI-powered practice between sessions.
This topic builds on GCSE scatter graphs. The earlier Statistics Year 1 topics, from Data Collection to Representations of Data, deal with one variable at a time; correlation looks at two variables together and asks whether, and how, they are related. At A-Level the emphasis is on interpretation: describing a relationship in context, judging what it does and does not show, and using a regression line sensibly.
Four skills make up this topic:
Scatter diagrams — plotting bivariate data, and identifying the explanatory and response variables
Describing correlation — positive, negative or none, strong or weak, always in the context of the question
Interpreting correlation critically — correlation does not imply causation, and distinct groups or outliers can make a scatter diagram misleading
Regression lines — interpreting the equation \(y = a + bx\) in context, and using it to make predictions, knowing when a prediction is reliable
Bivariate data and scatter diagrams
Bivariate data is data with two variables recorded for each item, such as the temperature and the number of drinks sold on each day. A scatter diagram plots each pair of values as one point, so you can see whether the variables are related.
The explanatory (or independent) variable is the one that is set or changes on its own, and it goes on the horizontal \(x\)-axis. The response (or dependent) variable is the one whose value depends on it, and it goes on the vertical \(y\)-axis. Ask yourself: "which variable could affect the other?"
Worked Example 1
For each pair of variables, state which is the explanatory variable and which is the response variable.(a) The number of hours a student spends revising and their test mark(b) The daily temperature and the number of iced drinks a café sells(c) The age of a car and its price
1
(a) Revising could affect the mark, but the mark cannot affect how long was spent revising. Explanatory: hours of revision; response: test mark
2
(b) The weather could affect sales, not the other way round. Explanatory: temperature; response: number of iced drinks sold
3
(c) A car's age could affect its price. Explanatory: age; response: price
Correlation describes how close the points on a scatter diagram are to lying on a straight line, and in which direction that line goes:
Positive correlation — as one variable increases, the other tends to increase
Negative correlation — as one variable increases, the other tends to decrease
No correlation — there is no linear relationship between the variables
The closer the points are to a straight line, the stronger the correlation; widely scattered points that still show a trend have weak correlation. Correlation measures a linear relationship only, so always describe what it means in context, for example "students who revised for longer tended to score higher marks".
Positive correlation
Negative correlation
No correlation
Worked Example 2
The scatter diagram below shows the number of hours, \(x\), that 12 students spent revising for a test and their test mark, \(y\). Describe the correlation and interpret it in context.
1
The points rise from left to right, so the correlation is positive
2
The points lie close to a straight line, so the correlation is strong
3
In context: the students who revised for longer tended to score higher marks
AnswerStrong positive correlation: students who spent more time revising tended to get higher marks
Scatter diagram for Worked Examples 2 and 7: hours of revision and test mark for 12 students. The points lie close to a rising straight line.
Correlation and causation
Correlation shows that two variables are associated; it does not show that a change in one causes a change in the other. Two variables can be strongly correlated because both depend on a third variable, sometimes called a confounding variable, or simply by coincidence. To argue that one variable causes another you need a reason in context, or a controlled experiment, not just a scatter diagram.
Worked Example 3
Over one summer, a seaside town records the number of ice creams sold and the number of people treated for sunburn each day. There is strong positive correlation between the two variables. A newspaper claims that eating ice cream causes sunburn. Comment on this claim.
1
Strong positive correlation means that on days when more ice creams were sold, more people tended to be treated for sunburn
2
There is no sensible reason in context why eating ice cream would cause sunburn
3
A third variable explains both: on hot, sunny days, more people buy ice cream and more people get sunburnt
AnswerThe claim is not justified: correlation does not imply causation. Both variables are likely to be caused by a third variable, hot sunny weather
Distinct sections of a population and outliers
Sometimes a scatter diagram contains two or more distinct groups of points, for example data from two different locations, species or products. Each group can have its own pattern, and describing all the points together can give a misleading picture. When you see separate clusters, describe the correlation within each group and explain how the groups differ.
A single point far from the pattern of the rest is an outlier. As in Representations of Data, decide whether it is an error, which should be corrected or removed, or a genuine but unusual value, which should normally be kept. An outlier can make the correlation look much weaker than it is for the rest of the data.
Worked Example 4
The scatter diagram below shows the age, \(x\) years, and the price, \(y\) (£1000s), of 18 second-hand cars: 9 of Model A, a small family car, and 9 of Model B, a luxury car.(a) Describe the correlation for all 18 cars.(b) Describe the correlation for each model separately.(c) Explain why the answer to (a) is misleading.
1
(a) Taken together, the points rise from the Model A cluster on the left to the Model B cluster on the right: positive correlation overall
2
(b) Within each model, the points fall from left to right and lie close to a straight line: strong negative correlation for Model A and for Model B. For each model, older cars tend to be cheaper
3
(c) The data comes from two distinct sections of the population. In this sample the Model B cars happen to be older, and they are more expensive because they are luxury cars, not because they are older
Answer(a) Positive overall(b) Strong negative within each model(c) The overall pattern comes from mixing two different models, and hides the fact that, for each model, price falls with age
Scatter diagram for Worked Example 4 and Q12: age and price of 18 cars. The Model A points (teal, lower left) and the Model B points (amber, upper right) each fall from left to right.
Regression lines
When a scatter diagram shows linear correlation, the relationship can be modelled by a regression line. The regression line of \(y\) on \(x\) is the line of best fit used to predict the response variable \(y\) from the explanatory variable \(x\). It is written
\[y = a + bx\]
where \(b\) is the gradient and \(a\) is the \(y\)-intercept. In context:
\(b\) is the change in \(y\) for each increase of 1 unit in \(x\), on average. If \(b > 0\) the correlation is positive; if \(b < 0\) it is negative
\(a\) is the predicted value of \(y\) when \(x = 0\). Only interpret \(a\) if \(x = 0\) is a sensible value within, or close to, the range of the data
📋 Exam board note
You use a regression line; you do not calculate it
In the Edexcel AS content, calculations to find a regression line from raw data are excluded: a question gives you the equation, and you interpret it and use it to make predictions. The product moment correlation coefficient and hypothesis tests for correlation will be covered in Statistics Year 2, whose topic pages are not yet published. Check your own board's specification for exactly where each of these appears.
Worked Example 5
A café records the daily temperature, \(x\) °C, and the number of iced drinks sold, \(y\), on 10 days. The temperatures range from 12 °C to 26 °C. The regression line of \(y\) on \(x\) is \(y = 3.40 + 3.55x\).(a) Interpret the value 3.55 in context.(b) Explain whether the value 3.40 has a sensible interpretation.
1
(a) 3.55 is the gradient \(b\): for each 1 °C rise in temperature, the café sells about 3.55 more iced drinks, on average
2
(b) 3.40 is the predicted number of drinks sold when \(x = 0\), that is at 0 °C
3
0 °C is well outside the data, which runs from 12 °C to 26 °C, so the model may not hold there and 3.40 should not be interpreted
Answer(a) About 3.55 more iced drinks are sold for each 1 °C rise in temperature(b) No, because 0 °C lies outside the range of the data
Scatter diagram for Worked Examples 5 and 6: temperature and iced drinks sold on 10 days, with the regression line \(y = 3.40 + 3.55x\) drawn across the range of the data, 12 °C to 26 °C.
Making predictions: interpolation and extrapolation
Substitute a value of \(x\) into the regression equation to predict \(y\). A prediction for an \(x\)-value inside the range of the data is interpolation and is usually reliable, especially when the correlation is strong. A prediction for an \(x\)-value outside the range is extrapolation and is unreliable, because there is no evidence that the linear pattern continues.
The regression line of \(y\) on \(x\) is designed to predict \(y\) from a given value of \(x\). Do not rearrange it to predict \(x\) from a given value of \(y\).
Worked Example 6
Use the regression line \(y = 3.40 + 3.55x\) from Worked Example 5 to estimate the number of iced drinks sold when the temperature is:(a) 19 °C(b) 35 °CComment on the reliability of each estimate.
1
(a) \(y = 3.40 + 3.55 \times 19 = 3.40 + 67.45 = 70.85\), so about 71 drinks
2
19 °C is inside the range 12 °C to 26 °C (interpolation), and the points lie close to the line, so this estimate is reliable
3
(b) \(y = 3.40 + 3.55 \times 35 = 3.40 + 124.25 = 127.65\), so about 128 drinks
4
35 °C is outside the range of the data (extrapolation), so this estimate is unreliable: the relationship may not stay linear at such high temperatures
Answer(a) About 71 drinks, reliable (interpolation)(b) About 128 drinks, unreliable (extrapolation)
Worked Example 7
For the revision data in Worked Example 2, where revision times range from 2 to 12 hours, the regression line of \(y\) on \(x\) is \(y = 31.1 + 4.11x\).(a) Estimate the mark of a student who revised for 7.5 hours.(b) Another student wants to score 75 and rearranges the equation to find how long to revise. Explain why this is not a valid use of the regression line.
7.5 hours is inside the range 2 to 12 hours, so this is interpolation and the estimate is reliable: about 62 marks
3
(b) The regression line of \(y\) on \(x\) is for predicting the mark \(y\) from a given revision time \(x\). Rearranging it to predict \(x\) from \(y\) uses it the wrong way round
4
Also, correlation does not show that revising for longer causes a higher mark, so no regression line can guarantee a mark of 75 from a set revision time
Answer(a) About 62 marks(b) The regression line of \(y\) on \(x\) should only be used to predict \(y\) from \(x\), not \(x\) from \(y\)
Exam tips
💡 Tip 1
Always describe correlation in context
"Positive correlation" on its own often earns only part of the marks. Add what it means for the variables in the question, for example "on warmer days, the café tended to sell more iced drinks".
💡 Tip 2
Interpret the gradient with units and "on average"
When interpreting \(b\) in \(y = a + bx\), say what a 1-unit increase in \(x\) is, what happens to \(y\), and include the units: "each extra 1 °C is associated with about 3.55 more drinks sold, on average".
💡 Tip 3
Check the range of the data before predicting
Before you comment on reliability, compare the \(x\)-value with the range of the data. Inside the range is interpolation (reliable if the correlation is strong); outside it is extrapolation (unreliable).
💡 Tip 4
Look for clusters and outliers
If a scatter diagram shows separate groups, describe each group separately and say why combining them could mislead. If one point is far from the rest, consider whether it is an error.
💡 Tip 5
Suggest a third variable when causation is claimed
To challenge a claim that one variable causes another, say that correlation does not imply causation and name a sensible third variable, in context, that could affect both.
Common mistakes
Common Mistake 1
Saying correlation proves causation
Strong correlation shows that two variables are associated, not that one causes the other. A third variable, or coincidence, may explain the pattern.
Common Mistake 2
Trusting a prediction outside the data
Extrapolating beyond the range of the \(x\)-values assumes the linear pattern continues, which may not be true. The same applies to interpreting \(a\) when \(x = 0\) is far outside the data.
Common Mistake 3
Using the regression line the wrong way round
The regression line of \(y\) on \(x\) predicts \(y\) from \(x\). Rearranging it to find \(x\) from a given \(y\) is not a valid use of the line.
Common Mistake 4
Describing correlation without context
Writing only "strong positive" leaves out what the relationship means. State what happens to one variable as the other increases, using the names of the variables.
Common Mistake 5
Treating distinct groups as one population
Combining two separate clusters can create, hide or even reverse a correlation. Describe each group on its own and explain the difference between them.
Practice questions
Work through each question before checking the worked solution. Give predictions from a regression line to a sensible degree of accuracy, and any value read from a diagram as an estimate.
Core Skills
Q1State the type of correlation you would expect between:(a) the outdoor temperature and a household's daily heating cost(b) the heights and shoe sizes of adults(c) the last digit of a person's phone number and their heightCore Skills
Q2For each pair, state the explanatory variable and the response variable:(a) the amount of fertiliser given to a plant and its height after 6 weeks(b) the number of people in a lift and the time it takes to reach the top floorCore Skills
Q3For a group of athletes, \(x\) is the number of hours of training per week, from 2 to 12 hours, and \(y\) is a fitness score. The regression line of \(y\) on \(x\) is \(y = 15 + 2.5x\).(a) Interpret the value 2.5 in context.(b) Estimate the fitness score of an athlete who trains for 8 hours a week.Core Skills
Q4A scatter diagram shows strong negative correlation between the number of hours of television students watch each week and their test marks.(a) Describe what this shows in context.(b) A teacher says this proves that watching television lowers test marks. Explain why the teacher is wrong.Core Skills
Scatter diagram for Q5, Q6 and Q15: arm span and height of 12 adults.
Q5The scatter diagram above shows the arm span and height of 12 adults.(a) Describe the correlation shown by the points, ignoring the one point that does not fit the pattern.(b) Write down the approximate coordinates of that point.Core Skills
Exam-Style
Q6For the arm span data in Q5:(a) give a reason why the point that does not fit the pattern is likely to be a recording error(b) state what should be done with this value(c) explain how removing it would change the strength of the correlationExam-Style
Q7For trees of one species aged between 5 and 40 years, the regression line of height, \(y\) m, on age, \(x\) years, is \(y = 1.2 + 0.35x\).(a) Interpret the value 0.35 in context.(b) Estimate the height of a tree aged 20 years.(c) Use the equation to estimate the height of a tree aged 100 years, and comment on the reliability of this estimate.Exam-Style
Scatter diagram for Q8: foot length and reading score for 10 children aged 6 (teal, lower left) and 10 children aged 11 (amber, upper right).
Q8The scatter diagram above shows the foot length and reading score of 20 children from a primary school: 10 aged 6 and 10 aged 11.(a) Describe the correlation for all 20 children.(b) Describe the correlation within each age group.(c) A parent concludes that children with bigger feet are better readers. Explain why this conclusion is wrong.Exam-Style
Q9For towns with between 5 and 60 cafés, there is strong positive correlation between the number of cafés, \(x\), and the number of road accidents in a year, \(y\). The regression line of \(y\) on \(x\) is \(y = 4 + 1.8x\).(a) Estimate the number of road accidents in a year in a town with 30 cafés.(b) A councillor says that if a town opened 10 more cafés, it would have 18 more accidents each year. Explain why this is not justified, and suggest a third variable that could explain the correlation.Exam-Style
Q10For a group of runners who train between 10 and 60 km per week, the regression line of 10 km race time, \(y\) minutes, on weekly training distance, \(x\) km, is \(y = 62.4 - 0.38x\).(a) State the type of correlation shown, and interpret \(-0.38\) in context.(b) Runner A trains 25 km per week and runner B trains 45 km per week. Estimate the difference between their race times.(c) Explain why the line should not be used for a runner who trains 80 km per week.Exam-Style
A* Challenge
Q11A company records its monthly advertising spend, \(x\) (£100s), and its monthly sales, \(y\) (£1000s), for months in which \(8 \le x \le 40\). The regression line of \(y\) on \(x\) passes through the points \((12, 47)\) and \((32, 83)\).(a) Find the equation of the regression line in the form \(y = a + bx\).(b) Interpret the value of \(b\) in context, in pounds.(c) The company plans to spend £5000 on advertising next month. Estimate its sales and comment on the reliability of your estimate.(d) Explain why the regression line does not show that increasing advertising will increase sales.A* Challenge
Q12For the cars in Worked Example 4, the regression lines of price, \(y\) (£1000s), on age, \(x\) years, are: Model A only (ages 1 to 5 years), \(y = 14.0 - 1.47x\); Model B only (ages 6 to 10 years), \(y = 44.1 - 2.35x\); all 18 cars together, \(y = 6.06 + 2.08x\).(a) Explain why the line for all 18 cars has a positive gradient although each model's price falls with age.(b) Estimate the price of a 4-year-old Model A car using the Model A line, and using the line for all 18 cars. Which estimate is more reliable? Give a reason.(c) Explain why neither the Model A line nor the Model B line should be used to estimate the price of a 4-year-old Model B car.A* Challenge
Q13For car journeys between 20 km and 400 km, the regression line of fuel used, \(y\) litres, on distance, \(x\) km, is \(y = 0.8 + 0.065x\).(a) Interpret 0.065 in context, and express it as fuel used per 100 km.(b) Fuel costs £1.48 per litre. Estimate the cost of the fuel for a 260 km journey.(c) A driver uses 30 litres of fuel on a journey and uses the equation to estimate the distance. Explain why this is not a valid use of the regression line.(d) Explain why the value 0.8 should be interpreted with caution.A* Challenge
Q14For taxi journeys between 1 and 15 miles, the regression line of fare, \(y\) (£), on distance, \(x\) miles, is \(y = 4.5 + 1.6x\). The taxi firm wants the equation with the fare, \(F\), in pence and the distance, \(d\), in kilometres. Use 1 mile \(=\) 1.6 km.(a) Show that \(F = 450 + 100d\).(b) Interpret the values 450 and 100 in context.(c) Estimate the fare for a 12 km journey, and for a 40 km journey. Comment on the reliability of each estimate.A* Challenge
Q15For the arm span data in Q5, with arm span \(x\) cm and height \(y\) cm, two regression lines of \(y\) on \(x\) are found: one using all 12 adults, and one with the outlier removed. The two equations are \(y = 53.0 + 0.707x\) and \(y = 13.6 + 0.922x\).(a) Decide which equation was found with the outlier removed, giving a reason based on the position of the outlier.(b) Estimate the height of an adult with an arm span of 175 cm using each equation.(c) Explain which estimate is more reliable.A* Challenge
Answers — full worked solutions
Core Skills (Q1–Q5)
Q1 — (a) Negative; (b) Positive; (c) None
1
(a) On warmer days less heating is needed, so the heating cost tends to fall as the temperature rises: negative correlation
2
(b) Taller adults tend to have bigger feet: positive correlation
3
(c) The last digit of a phone number has no connection with height: no correlation
(a) Negative correlation(b) Positive correlation(c) No correlation
Q2 — Explanatory: (a) Fertiliser, (b) Number of people
1
(a) The amount of fertiliser is chosen and could affect growth. Explanatory: amount of fertiliser; response: height of the plant
2
(b) The number of people could affect the time taken, not the other way round. Explanatory: number of people; response: time taken
(a) Explanatory: amount of fertiliser; response: height of the plant(b) Explanatory: number of people; response: time taken
Q3 — (a) 2.5 points per extra hour; (b) 35
1
(a) For each extra hour of training per week, the fitness score increases by about 2.5, on average
2
(b) \(y = 15 + 2.5 \times 8 = 35\). 8 hours is inside the range 2 to 12 hours, so this is interpolation
(a) The fitness score increases by about 2.5 for each extra hour of training per week, on average(b) \(y = 35\) (interpolation)
Q4 — (a) More TV, lower marks; (b) Correlation is not causation
1
(a) Students who watched more television each week tended to get lower test marks
2
(b) Correlation does not imply causation. The pattern may be explained by a third variable, for example the amount of time spent studying: students who study less may both watch more television and score lower marks
(a) Students who watched more television tended to get lower test marks(b) Correlation does not imply causation; a third variable, such as time spent studying, may explain the pattern
Q5 — (a) Strong positive; (b) About \((160, 190)\)
1
(a) The other 11 points rise from left to right and lie close to a straight line: strong positive correlation. Adults with longer arm spans tend to be taller
2
(b) The point that does not fit is at approximately \((160, 190)\): arm span about 160 cm, height about 190 cm. Accept readings within \(\pm 2\) cm on each axis
(a) Strong positive correlation(b) Approximately \((160, 190)\)
Exam-Style (Q6–Q10)
Q6 — Likely error; check and correct or remove; correlation becomes stronger
1
(a) For the other adults, height is close to arm span. A height of about 190 cm with an arm span of about 160 cm is a difference of about 30 cm, which is very unlikely, so the value was probably recorded wrongly
2
(b) Check the original record: if it is an error, correct it if possible, otherwise remove it from the data. If it turns out to be a genuine value, it should be kept
3
(c) The other 11 points lie close to a straight line, so removing the outlier would make the correlation stronger
(a) A 30 cm gap between height and arm span is very unlikely, so it was probably recorded wrongly(b) Check the record: correct or remove it if it is an error; keep it if it is genuine(c) The correlation would become stronger
Q7 — (a) 0.35 m per year; (b) 8.2 m; (c) 36.2 m, unreliable
1
(a) For each extra year of age, the height of a tree increases by about 0.35 m, on average
2
(b) \(y = 1.2 + 0.35 \times 20 = 1.2 + 7 = 8.2\) m. 20 years is inside the range 5 to 40 years
100 years is far outside the range 5 to 40 years, so this is extrapolation and is unreliable; trees do not keep growing at a constant rate, so the linear model is unlikely to hold
(a) About 0.35 m per extra year of age, on average(b) 8.2 m(c) 36.2 m, unreliable (extrapolation)
Q8 — (a) Strong positive; (b) None within each group; (c) Age affects both
1
(a) Taken together, the points rise from the age-6 cluster to the age-11 cluster: strong positive correlation overall
2
(b) Within each age group the points are scattered with no clear trend: no correlation (or very weak) between foot length and reading score for children of the same age
3
(c) The data comes from two distinct groups. Age is a third variable: older children have bigger feet and read better. Within an age group, foot length tells you nothing about reading, and correlation does not imply causation anyway
(a) Strong positive correlation overall(b) No correlation (or very weak) within each age group(c) Age is a third variable that affects both foot length and reading score
Q9 — (a) 58 accidents; (b) Correlation is not causation; e.g. population
1
(a) \(y = 4 + 1.8 \times 30 = 4 + 54 = 58\). 30 cafés is inside the range 5 to 60, so this is interpolation
2
(b) The gradient 1.8 describes how the number of accidents differs between towns with different numbers of cafés; it does not show that cafés cause accidents, so adding 10 cafés will not necessarily add \(1.8 \times 10 = 18\) accidents
3
A third variable such as the population (or size) of the town is likely to explain both: larger towns have more cafés and more traffic, and so more accidents
(a) 58 accidents(b) Correlation does not imply causation; a third variable, such as the population of the town, may explain both
Q10 — (a) Negative; (b) About 7.6 minutes; (c) Extrapolation
1
(a) The gradient is negative, so there is negative correlation. For each extra 1 km of training per week, the race time is about 0.38 minutes (about 23 seconds) shorter, on average
Difference \(= 52.9 - 45.3 = 7.6\) minutes, which is also \(0.38 \times (45 - 25) = 7.6\). Both distances are inside the range 10 to 60 km
4
(c) 80 km is outside the range 10 to 60 km, so using the line would be extrapolation; race times cannot keep falling at a constant rate as training increases, so the prediction would be unreliable
(a) Negative correlation; about 0.38 minutes shorter for each extra 1 km of training per week(b) About 7.6 minutes(c) 80 km is outside the range 10 to 60 km (extrapolation), so the prediction is unreliable
(a) Gradient \(b = \dfrac{83 - 47}{32 - 12} = \dfrac{36}{20} = 1.8\). Then \(47 = a + 1.8 \times 12\), so \(a = 47 - 21.6 = 25.4\). The equation is \(y = 25.4 + 1.8x\)
2
(b) An increase of 1 in \(x\) is an extra £100 of advertising, and an increase of 1.8 in \(y\) is an extra £1800 of sales. So each extra £100 spent on advertising is associated with about £1800 more sales, on average
3
(c) £5000 is \(x = 50\), so \(y = 25.4 + 1.8 \times 50 = 115.4\), that is sales of about £115 400
4
\(x = 50\) is outside the range \(8 \le x \le 40\), so this is extrapolation and the estimate is unreliable
5
(d) The regression line describes an association in past data; correlation does not imply causation. Sales may also depend on other variables, such as the time of year, which could affect both advertising spend and sales
(a) \(y = 25.4 + 1.8x\)(b) Each extra £100 spent on advertising is associated with about £1800 more sales, on average(c) About £115 400, unreliable (extrapolation)(d) Correlation does not imply causation; other variables may affect sales
Q12 — (b) £8120 (Model A line, more reliable) and £14 380
1
(a) The data comes from two distinct sections of the population. In this sample the Model B cars happen to be older (6 to 10 years), and they are more expensive because they are luxury cars, not because they are older, so when the two models are combined, price appears to rise with age, even though it falls with age within each model
2
(b) Model A line: \(y = 14.0 - 1.47 \times 4 = 14.0 - 5.88 = 8.12\), about £8120
3
Line for all 18 cars: \(y = 6.06 + 2.08 \times 4 = 6.06 + 8.32 = 14.38\), about £14 380
4
The Model A estimate is more reliable: 4 years is inside the Model A range (1 to 5 years), Model A shows strong correlation, and the combined line mixes two different models, so it does not describe how a Model A car's price changes with age
5
(c) The Model A line describes a different model, so it does not apply to a Model B car. For the Model B line, 4 years is outside its range of 6 to 10 years, so using it would be extrapolation
(a) The Model B cars are older and are more expensive luxury cars, so combining the models makes price appear to rise with age(b) Model A line: about £8120; line for all 18 cars: about £14 380. The Model A estimate is more reliable(c) The Model A line does not apply to a Model B car, and 4 years is outside the Model B range (extrapolation)
Q13 — (a) 6.5 litres per 100 km; (b) About £26.20
1
(a) Each extra kilometre uses about 0.065 litres more fuel, on average. Per 100 km: \(0.065 \times 100 = 6.5\) litres
2
(b) \(y = 0.8 + 0.065 \times 260 = 0.8 + 16.9 = 17.7\) litres. 260 km is inside the range 20 to 400 km, so this is interpolation
3
Cost \(= 17.7 \times 1.48 = 26.196\), so about £26.20
4
(c) The regression line of \(y\) on \(x\) is for predicting fuel used from distance. Using it to predict distance from fuel used is using it the wrong way round
5
(d) 0.8 is the predicted fuel used for a journey of 0 km. \(x = 0\) is outside the range 20 to 400 km, so the model may not hold there
(a) About 0.065 litres per extra kilometre; 6.5 litres per 100 km(b) About £26.20(c) The line of \(y\) on \(x\) should only be used to predict fuel used from distance(d) 0.8 is the predicted fuel used for 0 km, which is outside the range of the data
Q14 — (c) £16.50, reliable; £44.50, unreliable
1
(a) The fare in pence is \(F = 100y\), and \(d = 1.6x\), so \(x = \dfrac{d}{1.6}\)
(b) 100: each extra kilometre adds about 100 pence (£1) to the fare, on average. 450: a fixed charge of about 450 pence (£4.50); \(d = 0\) lies just outside the data (the shortest journeys are 1 mile, or 1.6 km), so this is only an estimate, but a fixed starting charge is sensible in context
4
(c) 12 km: \(F = 450 + 100 \times 12 = 1650\) pence \(=\) £16.50. 12 km is 7.5 miles, inside the range 1 to 15 miles, so this is reliable
5
40 km: \(F = 450 + 100 \times 40 = 4450\) pence \(=\) £44.50. 40 km is 25 miles, outside the range 1 to 15 miles, so this is extrapolation and unreliable
(a) \(F = 450 + 100d\), as required(b) 450: a fixed charge of about £4.50; 100: about £1 for each extra kilometre(c) 12 km: £16.50, reliable; 40 km: £44.50, unreliable (extrapolation)
Q15 — (a) \(y = 13.6 + 0.922x\); (b) 176.7 cm and 175.0 cm
1
(a) The outlier, at about \((160, 190)\), has a short arm span but lies far above the pattern. Including it pulls the left-hand end of the line up, which makes the line less steep and raises its intercept
2
So \(y = 53.0 + 0.707x\) (smaller gradient, larger intercept) includes the outlier, and \(y = 13.6 + 0.922x\) was found with the outlier removed
3
(b) With the outlier: \(y = 53.0 + 0.707 \times 175 = 53.0 + 123.725 = 176.725\), about 176.7 cm. Without the outlier: \(y = 13.6 + 0.922 \times 175 = 13.6 + 161.35 = 174.95\), about 175.0 cm
4
(c) The estimate from \(y = 13.6 + 0.922x\) (about 175.0 cm) is more reliable: the outlier is probably a recording error, the remaining points show strong correlation, and 175 cm is inside the range of the data, so this is interpolation
(a) \(y = 13.6 + 0.922x\)(b) With the outlier: about 176.7 cm; without the outlier: about 175.0 cm(c) The estimate from \(y = 13.6 + 0.922x\) (about 175.0 cm): it excludes the likely error and is interpolation
Summary
Bivariate data has two variables for each item. Put the explanatory (independent) variable on the \(x\)-axis and the response (dependent) variable on the \(y\)-axis.
Describe correlation as positive, negative or none, and as strong or weak, and always say what it means in context. Correlation measures a linear relationship.
Correlation does not imply causation: a third variable, or coincidence, may explain it. Distinct groups and outliers can make a scatter diagram misleading.
In the regression line of \(y\) on \(x\), \(y = a + bx\), \(b\) is the change in \(y\) per unit increase in \(x\), on average; only interpret \(a\) if \(x = 0\) is within, or close to, the range of the data.
Interpolation (inside the range of the data) is reliable when correlation is strong; extrapolation (outside the range) is unreliable. Use the line only to predict \(y\) from \(x\).
If interpreting scatter diagrams or regression lines is still causing problems, Alamin's diagnostic approach identifies exactly which skills are missing and builds a targeted plan to address them — with AI-powered practice between sessions.