4.4 Correlation and regression of y on x
Studying this course? Keep your progress in one place.
Key facts
- Use technology for and for the line .
- Describe correlation by direction (positive or negative) and strength (strong or weak), in context.
- measures linear correlation only; correlation does not imply causation.
- The line passes through .
- Interpret as the change per unit of and as the value at .
- Predict from only, and avoid extrapolation.
Scatter diagrams
A scatter diagram of paired data shows the direction and strength of any linear pattern.
Bivariate data are pairs measured on the same items. Positive correlation: tends to increase with . Negative correlation: tends to decrease. Zero correlation: no linear pattern. It is strong when points lie close to a line, weak when scattered.
A line of best fit by eye should pass through the mean point with roughly equal numbers of points on each side.
As the temperature rises, fewer people buy hot soup. What type of correlation is this?
Pearson's r
lies between and ; its sign gives the direction and its size the strength of the linear relationship.
Pearson's product-moment correlation coefficient is found with technology. near is strong positive; near strong negative; near little or no linear correlation. When a critical value is given, the correlation is significant if is larger than it.
Points on a curve can give a small even though and are closely related.
Strong
Moderate
Weak
Little or none
Which value of shows the strongest linear relationship?
Correlation is not causation
A strong correlation shows two variables move together, not that one causes the other.
The link may come from a third variable affecting both: hot weather increases both ice cream sales and sunburn. Or it may be a coincidence. In an exam comment, say correlation does not imply causation and name a plausible third factor in context.
- 1
Hot weather
the third variable
- 2
Affects both
more ice cream bought, more sunburn
- 3
Ice cream sales and sunburn rise together
strong correlation
- 4
Not causation
ice cream does not cause sunburn
Ice cream sales and sunburn cases are strongly correlated. Which is the best comment?
The regression line of y on x
The line is found with technology, passes through , and its coefficients have meanings in context.
It minimises the sum of the squared vertical distances from the points. The gradient is the change in for each increase of 1 in ; the -intercept is the predicted at , meaningful only if makes sense and is near the data.
For price against age, : the price falls by about 1850 dollars per year of age. Use unrounded values for later predictions.
Worked example
A car price (dollars) against age in years has regression line . Predict the price at 4 years and interpret the gradient.
- 1
Substitute : .
- 2
The gradient means the price falls by about 1850 dollars for each extra year of age.
In , what does 21400 represent?
Predicting safely
Interpolation inside the data range is reliable for strong correlation; extrapolation outside it is not.
A line predicting a revision score of 130 out of 100 shows why extrapolation fails. A on line should only predict from : rearranging it to predict is not valid. A prediction is also unreliable if is small.
Interpolation
- inside the data range
- Reliable if correlation is strong
Extrapolation
- outside the data range
- Unreliable: the pattern may not continue
A regression line is based on data with from 10 to 60. Which prediction is least reliable?
Try an exam question
The regression line of mark on revision hours is , based on data with between 2 and 10. (a) Interpret the gradient. (b) Predict the mark for . (c) Comment on using the line for .
[3 marks]
- [1]Each extra hour of revision raises the predicted mark by 2.4.
- [1].
- [1] is outside the data range (extrapolation), so the prediction is unreliable.
That's the notes covered.
Carry on to the next subtopic.
Try a question
Use the regression line to estimate the selling price of a 5-year-old car.
1 mark, marked against the mark scheme
Use the regression line to estimate the selling price of a 5-year-old car.
1 mark, marked against the mark scheme