Sign in

← All revision topics

English variety

Correlation versus causation; interpolation and extrapolation; regression line using technology

Machine translation

Around age 14 · Usually taught in ⁨Key Stage 4⁩ · Free

Learn

  1. If two variables are correlated, does one cause the other? Not necessarily. Mara records how many ice creams she sells and how many people walk along the canal (in tens). The correlation is strong, but the ice creams do not attract the people: both go up when it is hot. The heat is a third variable that has not been measured.

    12345678910123456789100Ice creamsPeople
    Strong correlation, but no direct cause
  2. When there is correlation, there are four possible explanations: that A causes B, that B causes A, that a third variable causes both, or that it is a coincidence. To prove a cause you need, for example, an experiment with a control group.

    Technology (a spreadsheet, a calculator or GeoGebra) finds the regression line, the line y = mx + b that best fits the scatter of points, and the correlation coefficient r. Using some friends' study hours and their marks gives y = 0.5x + 3 and r = 0.89.

    12345678910123456789100HoursMark
    Regression line: y = 0.5x + 3, with r = 0.89
  3. The gradient, 0.5, tells you how much y changes, on average, for each extra unit of x: each extra hour, 0.5 marks more. The intercept, 3, is what the line estimates for x = 0. The number r goes from −1 to 1: its sign is the sign of the correlation and, the closer to 1 or to −1, the stronger it is.

    The regression line always passes through the mean point: (mean of x, mean of y). The difference between a real data value and what the line estimates is data − line: it is positive if the point is above the line.

    You use the line to estimate. Inside the range of the data (from 2 to 10 hours) it is interpolating, and it is fairly reliable. Outside it is extrapolating: risky, because the relationship may change and you get values that make no sense.

    HoursEstimated markReliable?
    66Yes: interpolate
    87Yes: interpolate
    2013No: extrapolate
    A mark of 13, with a maximum of 10, makes no sense
  4. Be careful with r: it only measures how close the points are to a straight line. With r = 0 there can still be a curved relationship. Changing units, from cm to m, changes the gradient but not r. And a single outlier can change the line and r a lot.

Worked examples

Example 1

Mara records the temperature at eight in the morning and the cups of hot chocolate she sells in an hour (from 2 to 10 °C). The spreadsheet gives y = −0.5x + 10 and r = −0.89. What does −0.5 mean? How many cups does it estimate for 6 °C?

12345678910123456789100°CCups
  1. The gradient, −0.5, tells you that for each extra degree, 0.5 fewer cups are sold, on average: one cup fewer every 2 degrees.
  2. For 6 °C: y = −0.5 × 6 + 10 = −3 + 10 = 7.
  3. 6 °C is inside the data (from 2 to 10): it is interpolating and the estimate is reliable.

It estimates 7 cups. The gradient −0.5 is the average fall in cups for each extra degree.

Watch out! Here many people take the 10 for the gradient, or the −0.5 for the intercept.

Common mistake

Example 2

Gus sees that in the weeks when Mara sells more umbrellas there are more traffic jams. The correlation is strong. Can he say that the umbrellas cause the traffic jams?

  1. There is a strong positive correlation: both go up at the same time.
  2. But a correlation does not prove a cause: you have to think about other explanations.
  3. A third variable explains both: the rain makes people buy umbrellas and fills the streets with cars.

No: the rain explains both variables. The umbrellas do not cause the traffic jams.

Watch out! Here many people conclude that the umbrellas cause the traffic jams because the correlation is strong.

Common mistake

Example 3

With the data from 2 to 10 hours of study, the line is y = 0.5x + 3. What mark does it estimate for someone who studies 20 hours? Is it reliable?

HoursEstimated markReliable?
2013No: extrapolate
  1. y = 0.5 × 20 + 3 = 10 + 3 = 13.
  2. 20 hours is outside the range of the data (from 2 to 10): it is extrapolating.
  3. A mark of 13 is not possible if the maximum is 10: the estimate is not reliable.

The line estimates 13, but it is not reliable: it is an extrapolation.

Watch out! Here many people say they will get 13 because the line gives it.

Common mistake

A challenge to think about

Find a headline in a newspaper or on the internet that says one thing causes another. Which two variables does it link? Can it be explained by a third variable, or by the cause being the other way round? Write your explanation in three sentences.

Practise

Rounds of 10 questions. Choose a level:

Your grade in this topic:No grade yetDo a round at the Basic level to get a grade.