Scatter diagrams, correlation and regression

Scatter diagrams, correlation and regression

Spec L2 · AQA A-Level Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • Correlation is positive, negative or none; the closer to a line, the stronger.
  • In y=a+bxy=a+bx, the gradient bb is the average change in yy per 1 unit of xx.
  • The intercept aa is the predicted yy at x=0x=0; it only makes sense if x=0x=0 is in range.
  • Interpolation (inside the data range) is reliable; extrapolation is not.
  • Correlation does not imply causation: look for a third variable.

Correlation

A scatter diagram plots paired values. Correlation describes the direction and strength of the pattern.

A scatter diagram plots one point per item. Correlation is positive, negative or none, and the closer the points lie to a straight line, the stronger it is.

Describe both direction and strength, in context: 'the more hours of revision, the higher the test score tends to be'.

12345678910246810HoursScoreLine of best fit
Positive correlation (illustrative data): the more hours of revision, the higher the test score tends to be.

Positive

  • As one increases, the other tends to increase

Negative

  • As one increases, the other tends to decrease

No correlation

  • No clear pattern

Points on a scatter diagram lie very close to a line sloping downwards. How do you describe the correlation?

Regression lines

A regression line y=a+bxy=a+bx is the straight line of best fit. The gradient and intercept must be interpreted in context.

In y=a+bxy=a+bx, yy is the response variable and xx the explanatory variable. You will not be asked to calculate the line, only to interpret it.

The gradient bb is the average change in yy per increase of 1 in xx. The intercept aa is the predicted yy when x=0x=0, meaningful only if x=0x=0 is in the range of the data.

For s=34+4.5hs=34+4.5h, each extra hour of revision is associated with 4.5 more marks on average.

246810121420406080100xy(0, 34)(8, 70)s = 34 + 4.5h
Regression line s = 34 + 4.5h for revision hours h and test mark s

Worked example

Use s = 34 + 4.5h to estimate the mark after 8 hours of revision.

In a regression line y = 12 + 3x, what does the 3 mean?

Interpolation and extrapolation

Estimating inside the range of the data is interpolation and is reasonably reliable. Estimating outside it is extrapolation and is unreliable.

Interpolation is usually reliable if the correlation is strong. Extrapolation is unreliable because the relationship may not continue in the same way.

When asked whether an estimate is reliable, state the range of the data, say whether the value is inside or outside it, and comment on the strength of the correlation.

5101520253020406080100120140160180xyMaximum mark 100End of data(12, 88)(30, 169)s = 34 + 4.5h
Data cover h = 1 to 12. Extrapolating to h = 30 gives 169, impossible for a test marked out of 100.

Data cover h=1h=1 to 1212. Which estimate is most reliable?

Separate groups

Separate clusters often mean different groups. One line through all of them can mislead, so analyse each group separately.

The data may contain groups such as women and men, or weekdays and weekends. A single regression line through all the points can mislead: the overall correlation may come mainly from the difference between groups, and the line may fit neither group well.

Fit a separate line to each identifiable group, and compare gradients: a larger gradient means a steeper rise in yy per unit of xx.

12345678910246810xyOne lineGroup AGroup B
Illustrative data: one line through everything (dashed) rises, yet within each group the trend falls.

One line for everything

  • May fit neither group
  • Correlation may come from the gap between groups

A line for each group

  • Fits each group better
  • Compare gradients to compare groups

A scatter diagram shows two separate clusters. What is the best comment?

Causation

No. Two variables can be correlated without one causing the other: say the data show an association only.

Possible reasons for a correlation without causation:

  • A confounding variable (a third variable) drives both.
  • The link is a coincidence.
  • The direction of causation is the other way round.

To comment on a claim of causation, say the data show an association only, suggest another factor, and describe what extra evidence would help.

correlated, not causalHot weather (third variable)Ice-cream salesBeach rescues
A third variable can drive both: hot weather raises ice-cream sales and beach rescues, which are correlated but do not cause each other.

Third variable

  • Hot weather raises both ice-cream sales and beach rescues

Coincidence

  • The link is chance

Reverse direction

  • Y causes X, not X causes Y

Shoe size and reading ability are positively correlated in primary school children. What is the likely explanation?

Try an exam question

A regression line for ice creams sold yy against temperature tt (∘^\circC) is y=−40+12ty=-40+12t, from data with tt between 1515 and 3030. Interpret the gradient, estimate yy when t=24t=24, and explain why the line is unreliable for t=5t=5.

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Which statement about the value 4.5 in the regression line is correct?

1 mark, marked against the mark scheme

Revision notes